<?xml version="1.0" encoding="utf-8"?>

<?xml-stylesheet type="text/css"
  href="style/rss.css"?>

<!DOCTYPE rss
 PUBLIC "-//IDN Silmaril.IE/DTD RSS 2.0//EN"
 "http://WWW.Silmaril.IE/software/rss2.dtd">

<rss version="2.0">
  <channel>
    <title>sabi notes</title>
    <link>http://WWW.sabi.co.UK/</link>
    <description>Notes mostly about Linux and computer issues</description>
    <language>en-GB</language>

    <docs>http://blogs.law.Harvard.edu/tech/rss/</docs>
    <webMaster>notes {at} notes.for.sabi.co.UK</webMaster>
    <copyright>Copyright 2006 P. Grandi</copyright>
    <ttl>14000</ttl>

    <category>AMD</category>
    <category>Intel</category>
    <category>Microsoft</category>
    <category>Nintendo</category>
    <category>Sony</category>

    <category>CPU</category>
    <category>OS</category>
    <category>VoIP</category>
    <category>file sytem</category>
    <category>free software</category>
    <category>hardware</category>
    <category>performance</category>
    <category>system administration</category>
    <category>videogames</category>
    <category>virtual memory</category>
    
    <category>ALSA</category>
    <category>Debian</category>
    <category>Fedora</category>
    <category>GNU</category>
    <category>JFS</category>
    <category>KDE</category>
    <category>Linux</category>
    <category>MS Windows 2000</category>
    <category>MS Windows XP</category>
    <category>MS Windows</category>
    <category>PS/3</category>
    <category>RedHat</category>
    <category>SUSE</category>
    <category>UNIX</category>
    <category>XFS</category>
    <category>Xbox 360</category>
    <category>ext3</category>

    <lastBuildDate>Mon, 22 Jan 2007 21:05:00 +0000</lastBuildDate>

<!-- in progress items above -->

    <item>
      <title>Crazily high cable prices</title>
      <link>http://WWW.sabi.co.UK/blog/0701jan.html#070122</link>

      <description>Spotted impressively high prices for ordinary
	computer cables. Online too.</description>

      <category>cables</category>
      <category>shopping</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/blog/0701jan.html#070120b</guid>
      <pubDate>Mon, 22 Jan 2007 21:00:00 +0000</pubDate>
    </item>

    <item>
      <title>Simple configuration-driving environment variables</title>
      <link>http://WWW.sabi.co.UK/blog/0701jan.html#070120b</link>

      <description>A few environment variables can be a nice
	orthogonal base for configuration space, especially
	shell-script configuration files.</description>

      <category>UNIX</category>
      <category>Linux</category>
      <category>system administration</category>
      <category>shell</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/blog/0701jan.html#070120b</guid>
      <pubDate>Sat, 20 Jan 2007 23:15:00 +0000</pubDate>
    </item>

    <item>
      <title>Impressive FireWire 800 performance</title>
      <link>http://WWW.sabi.co.UK/blog/0701jan.html#070120</link>

      <description>Amazing performance of FireWire 800, where the same
	disk runs 2-3 times as fast with it than with FireWire 400 or
	USB2, and at speeds near those of SATA.</description>

      <category>storage</category>
      <category>hardware</category>
      <category>USB2</category>
      <category>SATA</category>
      <category>USB2</category>
      <category>IEEE1394</category>
      <category>FireWire</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/blog/0701jan.html#070120</guid>
      <pubDate>Sat, 20 Jan 2007 09:55:00 +0000</pubDate>
    </item>

    <item>
      <title>Trying two recent branded external hard drives</title>
      <link>http://WWW.sabi.co.UK/blog/0701jan.html#070115</link>

      <description>I have just tried two branded external USB2 hard
	drives, and they seem to work fine, and fairly quickly, and
	without the power supply issues that come from underspecified
	power bricks. But eSATA is going to be hopefully much
	better.</description>

      <category>storage</category>
      <category>hardware</category>
      <category>USB2</category>
      <category>SATA</category>
      <category>USB2</category>
      <category>IEEE1394</category>
      <category>FireWire</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/blog/0701jan.html#070115</guid>
      <pubDate>Mon, 15 Jan 2007 22:25:00 +0000</pubDate>
    </item>

    <item>
      <title>Parallel system with low power multi-CPU chips</title>
      <link>http://WWW.sabi.co.UK/blog/0701jan.html#070108</link>

      <description>SiCortex has launched massively parallel low power
	systems based in 6-CPU MIPS chips based on the idea that low
	power is good and memory stalls are bad.</description>

      <category>architecture</category>
      <category>hardware</category>
      <category>parallel</category>
      <category>CPU</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/blog/0701jan.html#070108</guid>
      <pubDate>Mon, 08 Jan 2007 22:10:00 +0000</pubDate>
    </item>

    <item>
      <title>The DNS, interfaces, nodes, directories and naming schemes</title>
      <link>http://WWW.sabi.co.UK/blog/0701jan.html#070105</link>

      <description>The DNS is a hierarchical linearized scheme for
	naming interfaces, not computers or other nodes. But people
	don't think in terms of interfaces, and nodes with multiple
	interfaces on multiple subnets pose some naming issues. Some
	alternatives are briefly discussed.</description>

      <category>networking</category>
      <category>sysadm</category>
      <category>Internet</category>
      <category>DNS</category>
      <category>naming</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/blog/0701jan.html#070105</guid>
      <pubDate>Fri, 05 Jan 2007 21:45:00 +0000</pubDate>
    </item>

    <item>
      <title>Per-node servers and interposition, and the DNS</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061231</link>

      <description>Putting in per-node caching or proxy servers is a
	form of non-transparent static interposition, a technique which
	in its general form is rarely available but results in great
	flexibility. It is related to advising and 'become:' in
	programming languages, and to DNS service discovery
	standards.</description>

      <category>networking</category>
      <category>languages</category>
      <category>sysadm</category>
      <category>DNS</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061231</guid>
      <pubDate>Sun, 31 Dec 2006 23:20:00 +0000</pubDate>
    </item>

    <item>
      <title>Fighting with YumEx and Yum</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061230b</link>

      <description>Using YumEx and Yum to set up a laptop with a lot of
	tools, and suffering because of their slowness and limitations.
	Things could be better than this...</description>

      <category>RedHat</category>
      <category>Fedora</category>
      <category>performance</category>
      <category>Linux</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061230b</guid>
      <pubDate>Sat, 30 Dec 2006 20:10:00 +0000</pubDate>
    </item>

    <item>
      <title>Amazing networking performance</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061230</link>

      <description>Some simple speed tests show that current top end
	CPUs and system can transmit at full speed on 1gb/s wires even
	without jumbo frames as they can sustain amazingly high interrupt
	and context switch rates, and intelligent network adapters
	help.</description>

      <category>networking</category>
      <category>performance</category>
      <category>Linux</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061230</guid>
      <pubDate>Sat, 30 Dec 2006 11:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Difference between Smalltalk-72 and -76, why it matters</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061227</link>
      <description>Smalltalk-72, Smalltalk-76, the big difference
	between message passing and overloading why it matters,
	and many CPUs on the same chip.</description>
      <category>programming</category>
      <category>software engineering</category>
      <category>object oriented</category>
      <category>overloading</category>
      <category>message passing</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061227</guid>
      <pubDate>Wed, 27 Dec 2006 15:10:00 +0000</pubDate>
    </item>

    <item>
      <title>Rapid or less painful switching of service provider</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061226b</link>

      <description>It is often useful to run caching or slave servers on
	all clients so that most server addresses can be configured as
	127.0.0.1 as reconfiguring and restarting servers is often a lot
	quicker and less painful than reconfiguring and restarting
	applications.</description>

      <category>sysadm</category>
      <category>servers</category>
      <category>networking</category>
      <category>Internet</category>
      <category>Ethernet</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061226b</guid>
      <pubDate>Tue, 26 Dec 2006 21:55:00 +0000</pubDate>
    </item>

    <item>
      <title>Looking again at centralized vs. localized servers: why, where</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061226</link>

      <description>Centralized vs. localized service depends on a table
	of service vs. level of redundancy. Partitioning it into
	subtables and then reading it by rows or columns gives different
	service structures. Many services have dependencies, therefore
	it is pointless to have independent redundancy for them. In
	general the idea of bringing servers nearer to the point of
	delivery, usually a workgroup, is indeed reasonable according to
	some plausible criteria.</description>

      <category>sysadm</category>
      <category>servers</category>
      <category>networking</category>
      <category>Internet</category>
      <category>Ethernet</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061226</guid>
      <pubDate>Tue, 26 Dec 2006 14:40:00 +0000</pubDate>
    </item>

    <item>
      <title>Message passing and overloading</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061224</link>
      <description>Smalltalk-72, Smalltalk-76, the big difference
	between message passing and overloading, and terminology
	confusion.</description>
      <category>programming</category>
      <category>software engineering</category>
      <category>object oriented</category>
      <category>overloading</category>
      <category>message passing</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061224</guid>
      <pubDate>Sun, 24 Dec 2006 13:55:00 +0000</pubDate>
    </item>

    <item>
      <title>OO, overloading and genericity</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061223</link>
      <description>What are overloading and genericity and their
	relationship to OO.</description>
      <category>programming</category>
      <category>software engineering</category>
      <category>object oriented</category>
      <category>modules</category>
      <category>overloading</category>
      <category>genericity</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061223</guid>
      <pubDate>Sat, 23 Dec 2006 18:20:00 +0000</pubDate>
    </item>

    <item>
      <title>Tomcat 5 configuration for multiple ports</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061216</link>

      <description>Tomcat has a slightly odd way of resolving request
	URLs into servlets. Binding different ports on the same domain
	to different default servlets requires a somewhat unusual
	configuration, but it is useful for one-to-one URL forwarding
	from an Apache front-end via AJP.</description>

      <category>sysadm</category>
      <category>Linux</category>
      <category>Apache</category>
      <category>Tomcat</category>
      <category>web</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061216</guid>
      <pubDate>Sat, 16 Dec 2006 20:35:00 +0000</pubDate>
    </item>

    <item>
      <title>More on large bridged network and centralized servers</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061212</link>

      <description>More details about VLANs: bridged VLANs are often
	partitioned by VLAN tagging, two opposite but complementary
	techniques. These are often used together in large networks, which
	mostly means inter-site and even international ones, both to take
	advantage of large broadcast domains and mitigate their effects,
	especially when servers and services are
	centralized.</description>

      <category>networking</category>
      <category>sysadm</category>
      <category>Linux</category>
      <category>NFS</category>
      <category>VLAN</category>
      <category>Ethernet</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061212</guid>
      <pubDate>Tue, 12 Dec 2006 21:45:00 +0000</pubDate>
    </item>

    <item>
      <title>TCP: data reliable, signaling not reliable</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061210b</link>

      <description>Some application problems are explained by pointing
	out that while TCP transport is reliable, signaling is not, and
	anyhow TCP under Linux has a very long and excessive timeout for
	detecting a dead connection.</description>

      <category>networking</category>
      <category>Internet</category>
      <category>TCP</category>
      <category>NFS</category>
      <category>Linux</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061210b</guid>
      <pubDate>Sun, 10 Dec 2006 19:20:00 +0000</pubDate>
    </item>

    <item>
      <title>Large bridged networks with centralized servers</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061210</link>

      <description>Large network infrastructures based on bridging and
	central servers are often the result of hasty attempts to scale
	up smaller infrastructures. They may seem to work, but I think
	that they are fragile and inefficient. I reckon that workgroup
	servers and local services are preferable. Difficult choices
	anyhow about home directories and laptops.</description>

      <category>networking</category>
      <category>sysadm</category>
      <category>Linux</category>
      <category>NFS</category>
      <category>VLAN</category>
      <category>Ethernet</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061210</guid>
      <pubDate>Sun, 10 Dec 2006 17:35:00 +0000</pubDate>
    </item>

    <item>
      <title>Updated IP shaping script</title>
      <link>http://WWW.sabi.co.UK/#sourcesSabishape?061126</link>
      <description>
	While the logic of my traffic shaping setup script has not
	changed much, this update contains many adjustments to details
	that should improve it a fair bit.
      </description>
      <category>Linux</category>
      <category>networking</category>
      <category>Netfilter</category>
      <category>iptables</category>
      <category>shell</category>
      <category>shaping</category>
      <category>IP</category>
      <category>configuration</category>
      <category>traffic control</category>

      <pubDate>Sun, 26 Nov 2006 15:30:00 +0000</pubDate>
    </item>

    <item>
      <title>How long does it take for a standard PC to become fatally infected</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061111</link>

      <description>An article reports that after a rather short time of
	being connected to the Internet and browsing an unprotected
	installation of MS Windows XP gets infected with very many
	trojans, spyware, ... programs.</description>

      <category>Microsoft</category>
      <category>Internet</category>
      <category>security</category>
      <category>spyware</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061111</guid>
      <pubDate>Sat, 11 Nov 2006 22:05:00 +0000</pubDate>
    </item>

    <item>
      <title>Impressions from LinuxWorld Expo 2006, London</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061105</link>

      <description>Belated impressions from the recent Linux World Expo
	in London.</description>

      <category>Linux</category>
      <category>IBM</category>
      <category>RedHat</category>
      <category>Novell</category>
      <category>system administration</category>
      <category>network administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061105</guid>
      <pubDate>Sun, 05 Nov 2006 22:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Very high spyware and trojan infection rates</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061103f</link>

      <description>A study (from an interested party) reports very,
	very high rates of spyware and trojan infections. Owning
        someone else's computer can be quite profitable.</description>

      <category>security</category>
      <category>MS Windows</category>
      <category>privacy</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061103f</guid>
      <pubDate>Fri, 03 Nov 2006 22:10:00 +0100</pubDate>
    </item>

    <item>
      <title>FireWire over Ethernet style wiring</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061103e</link>

      <description>Just discovered a standard to run the FireWire bus
	protocol over 1000gb/s Ethernet wiring. Quite an interesting
	combination.</description>

      <category>cabling</category>
      <category>FireWire</category>
      <category>Ethernet</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061103e</guid>
      <pubDate>Fri, 03 Nov 2006 21:55:00 +0100</pubDate>
    </item>

    <item>
      <title>Outsourcing incentives to stagnation</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061103d</link>

      <description>A pundit points out that outsourcing often results in
	high incentives to stagnation for the outsourcing
	supplier.</description>

      <category>sociology</category>
      <category>business</category>
      <category>outsourcing</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061103d</guid>
      <pubDate>Fri, 03 Nov 2006 21:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Lots of lost laptops</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061103c</link>

      <description>An article reports amazing numbers of lost laptops,
	of which few are reclaimed, because managers prefer to get a new
	new than wasting time in recovering the old one.</description>

      <category>security</category>
      <category>sociology</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061103c</guid>
      <pubDate>Fri, 03 Nov 2006 21:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Shoddy mechanical design for connectors</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061103b</link>

      <description>Some connector designs are ptrone to breaking off,
	and this is bad particularly for sockets. FireWire connectors
	were designed well though.</description>

      <category>design</category>
      <category>quality</category>
      <category>mechanical</category>
      <category>connector</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061103b</guid>
      <pubDate>Fri, 03 Nov 2006 21:20:00 +0100</pubDate>
    </item>

    <item>
      <title>John C. Dvorak on CSS</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061103</link>

      <description>A pundit points out that CSS is a very dense and
	compact languages, and mishaps like poor implementation have
	wide effects.</description>

      <category>web</category>
      <category>CSS</category>
      <category>HTML</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061103</guid>
      <pubDate>Fri, 03 Nov 2006 21:05:00 +0100</pubDate>
    </item>

    <item>
      <title>Important details to consider as to digital cameras</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061101</link>

      <description>Some small attributes of digital cameras are actually
	quite relevant, like having an external flash socket.</description>

      <category>photography</category>
      <category>digital</category>
      <category>camera</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061101</guid>
      <pubDate>Wed, 01 Nov 2006 21:40:00 +0100</pubDate>
    </item>


    <item>
      <title>EMC2 often recommends RAID3</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061031b</link>

      <description>Just found out that often EMC2, a noted storage
	system manufacturer, defaults to RAID3 or a variant.</description>

      <category>RAID</category>
      <category>storage</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061031b</guid>
      <pubDate>Tue, 31 Oct 2006 21:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Storage wire and command protocols, and SAN vs. NAS</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061031</link>

      <description>Commenting on a discussion about the difference
	between the two levels of storage protocols (link and command)
	and why bother with SANs as NAS is possible.</description>

      <category>NAS</category>
      <category>SAN</category>
      <category>SCSI</category>
      <category>ATA</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061031</guid>
      <pubDate>Tue, 31 Oct 2006 20:15:00 +0100</pubDate>
    </item>

    <item>
      <title>Teknik posture chair review</title>
      <link>http://WWW.sabi.co.UK/Notes/hwNotesReviews.html#teknikPosture?061022</link>
      <description>Using a "kneeling chair" to site somewhat high while
	not putting weight on the thighs.</description>
      <category>hardware</category>
      <category>seating</category>
      <category>ergonomics</category>

      <pubDate>Sun, 22 Oct 2006 18:40:00 +0100</pubDate>
    </item>

    <item>
      <title>File system notes updates</title>
      <link>http://WWW.sabi.co.UK/Notes/linuxFS.html?061022</link>

      <description>Some updates to the bibliography section and the
	links section of my file system for Linux notes.</description>

      <category>Linux</category>
      <category>RAID</category>
      <category>XFS</category>
      <category>JFS</category>
      <category>file system</category>

      <pubDate>Sun, 22 Oct 2006 17:45:00 +0100</pubDate>
    </item>

    <item>
      <title>RAID5 perversions</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061022b</link>

      <description>A list of quotes showing that some people misuse
	RAID5 as in 15+1 disc drive configurations, or worse.</description>

      <category>Linux</category>
      <category>RAID</category>
      <category>XFS</category>
      <category>file system</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061022b</guid>
      <pubDate>Sun, 22 Oct 2006 17:10:00 +0100</pubDate>
    </item>

    <item>
      <title>XFS etc. performance for parallel IO and fragmentation</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061022</link>

      <description>A pointer to and some comments on a very interesting
	paper showing some large recent improvements in XFS performance
	on a parallel system with multiple concurrent reads and writes.
	Some surprising comparisons with JFS and other file
	systems.</description>

      <category>Linux</category>
      <category>XFS</category>
      <category>JFS</category>
      <category>file system</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061022</guid>
      <pubDate>Sun, 22 Oct 2006 16:05:00 +0100</pubDate>
    </item>

    <item>
      <title>Options for mailboxing and tagged queueing</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061015</link>

      <description>I have recently had a look and upgraded a bit a small
	video-on-demand system that I first set-up about 10 years ago.
	Amusing to see how hardware had evolved in the meantime, and why
	I chose certain components may be interesting.</description>

      <category>architecture</category>
      <category>hardware</category>
      <category>video</category>
      <category>movies</category>
      <category>SCSI</category>
      <category>network</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061015</guid>
      <pubDate>Sun, 15 Oct 2006 11:50:00 +0100</pubDate>
    </item>


    <item>
      <title>Effect of elevator on multistream reading performance</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061014</link>

      <description>Some more experiments with multiple concurrent stream
	reading from a disc. Turns out that the 'anticipatory' elevator
	in Linux maintains aggregate bandwidth by sorting and batching
	requests, a bit cruder but overall with the same results as
	tagged queueing.</description>

      <category>architecture</category>
      <category>hardware</category>
      <category>video</category>
      <category>movies</category>
      <category>SCSI</category>
      <category>network</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061014</guid>
      <pubDate>Sat, 14 Oct 2006 08:55:00 +0100</pubDate>
    </item>

    <item>
      <title>Evolution of a video-on-demand system</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061013</link>

      <description>I have recently had a look and upgraded a bit a small
	video-on-demand system that I first set-up about 10 years ago.
	Amusing to see how hardware had evolved in the meantime, and why
	I chose certain components may be interesting.</description>

      <category>architecture</category>
      <category>hardware</category>
      <category>video</category>
      <category>movies</category>
      <category>SCSI</category>
      <category>network</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061013</guid>
      <pubDate>Fri, 13 Oct 2006 17:10:00 +0100</pubDate>
    </item>


    <item>
      <title>Another technique called "HDR" somewhat improperly</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061004</link>

      <description>In video games HDR rendering means doing intermediate
	lighting calculations with higher range than that used for final
	output, to prevent clipping due to overflow and underflow. This
	could be done by more careful staging of calculations, but it
	also delivers higher precision, not range, which helps with
	multipass algorithms</description>

      <category>architecture</category>
      <category>hardware</category>
      <category>graphics</category>
      <category>videogames</category>
      <category>NVIDIA</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061004</guid>
      <pubDate>Wed, 04 Oct 2006 19:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Games HDR, photos range compressions, and terminology</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-4th.html#061003</link>

      <description>Discovered High Dynamic Range photography, which is
	interesting but misnomed, unlike High Dynamic Range in
	videogames, which is truer to its name.</description>

      <category>architecture</category>
      <category>hardware</category>
      <category>graphics</category>
      <category>videogames</category>
      <category>NVIDIA</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-4th.html#061003</guid>
      <pubDate>Tue, 03 Oct 2006 23:45:00 +0100</pubDate>
    </item>

    <item>
      <title>Intel researching chip with 80 FPUs</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060928c</link>

      <description>Intel has been talking about a very fast chip with 80
	FPUs to be delivered in about 5 years. There is also already
	available the ClearSpeed CSX600 with 96 FPUs, but apparently it
	is much more expensive.</description>

      <category>architecture</category>
      <category>hardware</category>
      <category>CPU</category>
      <category>Intel</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060928c</guid>
      <pubDate>Thu, 28 Sep 2006 22:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates to the hardware notes and reviews</title>
      <link>http://WWW.sabi.co.UK/Notes/hwNotesReviews.html#belkinF5D7630?060928</link>
      <description>Updated the review of the Belkin F5D7630 with further
	condemnation of the poor quality of its firmware, especially
	after loading on it the much better firmware of the compatible
	SMC 7804WBRA.</description>

      <category>ADSL</category>
      <category>quality</category>
      <category>hardware</category>
      <category>networking</category>
      <category>system administration</category>

      <pubDate>Thu, 28 Sep 2006 18:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Fixing Belkin ADSL gateway bugs by using SMC firmware</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060928b</link>

      <description>My Belkin F5D7630 ADSL modem/router has always been
	buggy, but I managed to work around most bugs. Being fed up with
	it, I loaded on it the firmware from the compatible SMC 7804WBRA
	and what a difference. It seems clear to me that while Belkin
	and SMC used the same source, SMC spent some effort in testing
	and fixing the firmware, and documenting it well. Makes me wish
	I bought the SMC model.</description>

      <category>ADSL</category>
      <category>quality</category>
      <category>hardware</category>
      <category>networking</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060928b</guid>
      <pubDate>Thu, 28 Sep 2006 18:15:00 +0100</pubDate>
    </item>

    <item>
      <title>Configuring 'automount' to use '/etc/fstab'</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060928</link>

      <description>The Linux 'automount' daemon is quite convenient but
	is uses a different mount map format from '/etc/fstab'. But it
	can get the mount map by running a program, and it is was to
	write a script to convert '/etc/fstab' entries into 'automount'
	map ones on the line, and this preserves the flexibility and
	maintainability of keeping all mount configuration in one
	place.</description>

      <category>UNIX</category>
      <category>Linux</category>
      <category>filesystems</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060928</guid>
      <pubDate>Thu, 28 Sep 2006 18:00:00 +0100</pubDate>
    </item>

    <item>
      <title>Some reports on dual core versus hyperthreading performance</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060924c</link>

      <description>Spotted a couple on entries in the PostgreSQL mailing
	list comparing dual core AMD verbus hyperthreading Intel CPUs
	for running that highly parallel DBMS. AMD dual core looks good,
	but surprisingly so does hyperthreading.</description>

      <category>Linux</category>
      <category>Apache</category>
      <category>AMD</category>
      <category>Athlon</category>
      <category>Opteron</category>
      <category>performance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060924</guid>
      <pubDate>Sun, 24 Sep 2006 23:05:00 +0100</pubDate>
    </item>

    <item>
      <title>IBM to deliver hybrid Opteron and Cell cluster</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060924b</link>

      <description>IBM will deliver a 32,000 CPU Linux based cluster
	where half of the elements will be Opteron based and half Cell
	based, and claim that some new software infrastructure will help
	manage the cluster.</description>

      <category>Linux</category>
      <category>Opteron</category>
      <category>AMD</category>
      <category>IBM</category>
      <category>Cell</category>
      <category>cluster</category>
      <category>performance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060924b</guid>
      <pubDate>Sun, 24 Sep 2006 22:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Web transaction test shows 64b slower than 32b</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060924</link>

      <description>Some recent report shows that the same large scale
	web benchmark on system with 2 Opteron CPUs runs noticeable faster
	in 32b than 64b mode for between 128 and 100 and 400 clients.
	Comparing with a similar report for a test for 2 Athlon MP CPUs
	it looks like the reason is memory: the same program compiled
	for 32b mode uses less memory, and memory bandwidth, than in 64b
	mode.</description>

      <category>Linux</category>
      <category>Apache</category>
      <category>AMD</category>
      <category>Athlon</category>
      <category>Opteron</category>
      <category>performance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060924</guid>
      <pubDate>Sun, 24 Sep 2006 19:45:00 +0100</pubDate>
    </item>

    <item>
      <title>Polipo, another nice proxy cache for small sites</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060920b</link>

      <description>A pointer to Polipo, a nice and featureful proxy
	cache one might want to consider as an alternative to Apache or
	Squid, especially for IPv6 sites.</description>

      <category>Squid</category>
      <category>Apache</category>
      <category>programming</category>
      <category>Internet</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060920b</guid>
      <pubDate>Wed, 20 Sep 2006 23:05:00 +0100</pubDate>
    </item>

    <item>
      <title>Using Apache 2.2 as a proxy cache, with patch</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060920</link>

      <description>For various reasons I am now using Apache as a proxy
	cache instead of Squid. This has required fixing some bugs and
	extending a bit the Apache caching module.</description>

      <category>Squid</category>
      <category>Apache</category>
      <category>programming</category>
      <category>Internet</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060920</guid>
      <pubDate>Wed, 20 Sep 2006 08:40:00 +0100</pubDate>
    </item>

    <item>
     <title>IPv6 prefix non-portability and unique local addresses</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060917c</link>

      <description>According to one argument, local IPv6
	addresses have the advantage of being portable. But only
	in the sense of always requiring NAT to a non-portable prefix.
	An old non-portable global prefix can also be NAT'ed to
	a new different non-portable global prefix.</description>

      <category>network</category>
      <category>Internet</category>
      <category>IPv6</category>
      <category>NAT</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060917c</guid>
      <pubDate>Sun, 17 Sep 2006 23:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Reintroducing private addresses, and/or NAT, in IPv6</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060917b</link>

      <description>Some network administrators would like to have
	something like IPv4 private addresses and NAT in IPv6; a couple
	of ways have been proposed to achieve something like that, but
	things could be simpler, understanding what is the real goal,
	which is not private addresses or NAT as such.</description>

      <category>network</category>
      <category>Internet</category>
      <category>IPv6</category>
      <category>NAT</category>
      <category>security</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060917b</guid>
      <pubDate>Sun, 17 Sep 2006 18:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Weird and not so weird long standing bugs</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060917</link>

      <description>It is somewhat amazing that some minor but annoying
	bugs have remained in some very popular applications for several
	years. I haven't fixed them myself simply to avoid maintaining
	my own patches, and I haven't reported them to avoid the
	hassle.</description>

      <category>programming</category>
      <category>KDE</category>
      <category>Emacs</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060917</guid>
      <pubDate>Sun, 17 Sep 2006 17:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Large IPv6 datagram sizes and why they matter</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060916</link>

      <description>Why IPv6 larger datagram sizes matter when
	link bandwidths are high, not just window size when
	latency is also high.</description>

      <category>network</category>
      <category>Internet</category>
      <category>IPv6</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060916</guid>
      <pubDate>Sat, 16 Sep 2006 11:20:00 +0100</pubDate>
    </item>

    <item>
      <title>Thousands of queries per second for private addresses</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060915</link>

      <description>The root DNS servers get thousands of queries
	per second about site local private address ranges. As
	Stonebraker learned decades ago, misconfigured servers
	and networks are the norm, not the exception.</description>

      <category>network</category>
      <category>Internet</category>
      <category>DNS</category>
      <category>DBMS</category>
      <category>performance</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060915</guid>
      <pubDate>Fri, 15 Sep 2006 08:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Game load times, fragmentation; reporting to base</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060914b</link>

      <description>Using statistics collected from Steam users,
	Valve has discovered that poor game loading times depend
	a lot of fragmentation. Also means they scan users disks
	and report that back.</description>

      <category>videogames</category>
      <category>business</category>
      <category>performance</category>
      <category>file systems</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060914b</guid>
      <pubDate>Thu, 14 Sep 2006 19:25:00 +0100</pubDate>
    </item>

    <item>
      <title>Episodic game content four times cheaper to develop</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060914</link>

      <description>Valve reports that developing new episodes of
	an already developed game is four times cheaper. Also,
	that World of Warcraft is the benchmark.</description>

      <category>videogames</category>
      <category>business</category>
      <category>middleware</category>
      <category>MMORPG</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060914</guid>
      <pubDate>Thu, 14 Sep 2006 19:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Socket duty cycles matter</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060913b</link>

      <description>The designers of the original FireWire
	connectors designed them to be resistant to repeated
	plugging/unplugging. The number of insertion cycles for
	which sockets are rated is indeed in general an
	important and underrated topic.</description>

      <category>hardware</category>
      <category>FireWire</category>
      <category>IEEE1394</category>
      <category>security</category>
      <category>debugging</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060913b</guid>
      <pubDate>Wed, 13 Sep 2006 20:05:00 +0100</pubDate>
    </item>

    <item>
      <title>Astonishing FireWire security vulnerability</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060913</link>

      <description>Most FireWire host adapters offer excessive accessive
	access to their host's memory, and devices can initiate
	operations because FireWire is essentially P2P. This can be
	turned to advantage for debugging.</description>

      <category>hardware</category>
      <category>FireWire</category>
      <category>IEEE1394</category>
      <category>security</category>
      <category>debugging</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060913</guid>
      <pubDate>Wed, 13 Sep 2006 19:25:00 +0100</pubDate>
    </item>

    <item>
      <title>IPv6 options for IPv4 interoperability</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060911</link>

      <description>More on IPv6, the practical options to get IPv6 
	connectivity.</description>

      <category>network</category>
      <category>Internet</category>
      <category>IPv6</category>
      <category>security</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060911</guid>
      <pubDate>Mon, 11 Sep 2006 22:30:00 +0100</pubDate>
    </item>

    <item>
      <title>QoS shaping and a nice paper with a strange setup</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060910b</link>

      <description>Found a nice paper about QoS with Linux using the
	LARTC framework. Interesting observations about sampling rate.
	But puzzling setup of an example: incoming shaping is done at
	the gateway egree interface, not the ingress one.</description>

      <category>Linux</category>
      <category>LARTC</category>
      <category>network</category>
      <category>Internet</category>
      <category>QoS</category>
      <category>sabishape</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060910b</guid>
      <pubDate>Sun, 10 Sep 2006 20:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Using IPv6 for a large site</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060910</link>

      <description>While chatting about network management at a large
	scientific site IPv6 was mentioned. I quite like it, and anyhow it
	is going to become indispensable soon.  It also has some nice
	advantages by design, and some serendipitous ones.</description>

      <category>network</category>
      <category>Internet</category>
      <category>IPv6</category>
      <category>security</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060910</guid>
      <pubDate>Sun, 10 Sep 2006 18:05:00 +0100</pubDate>
    </item>

    <item>
      <title>Good value large can of grapefruit segments</title>
      <link>http://WWW.sabi.co.UK/Notes/miscShopping.html#princesGrapefruitLarge</link>

      <description>Just discovered my local Coop has a 1.25kg
	can of grapefruit segments for a good price. Excellent
	value.</description>

      <category>shopping</category>
      <category>food</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/miscShopping.html#princesGrapefruitLarge</guid>
      <pubDate>Mon, 11 Sep 2006 18:55:00 +0100</pubDate>
    </item>

    <item>
      <title>A new 'init' design from Ubuntu</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060828</link>

      <description>The Ubuntu project is going to replace the ugly
	System V style 'init' with 'upstart', which will be a general
	process manager, replacing 'cron' and 'inetd' as well. Seems
	a bit of a mess to me, but let's hope it is not.</description>

      <category>UNIX</category>
      <category>Linux</category>
      <category>pragmatics</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060828</guid>
      <pubDate>Mon, 28 Aug 2006 23:45:00 +0100</pubDate>
    </item>

    <item>
      <title>Much better performance with direct HyperTransport interface</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060825</link>

      <description>There are now motherboards with a direct
	HyperTransport interface called HTX, and performance seems to be
	much higher than PCI-E for high traffic network and probably
	most other coprocessors.</description>
      <category>hardware</category>
      <category>CPU</category>
      <category>AMD</category>
      <category>Intel</category>
      <category>bus</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060825</guid>
      <pubDate>Fri, 25 Aug 2006 22:40:00 +0100</pubDate>
    </item>

    <item>
      <title>AMD hopes for large market share in servers</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060822</link>
      <description>AMD hopes for a 40% market share in servers, where it
	has the advantage because of HyperTransport. Intel instead
	dominates laptops, especially with Core 2 Duo, and low end
	desktops with integrated graphics.</description>
      <category>hardware</category>
      <category>CPU</category>
      <category>AMD</category>
      <category>Intel</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060822</guid>
      <pubDate>Tue, 22 Aug 2006 22:05:00 +0100</pubDate>
    </item>

    <item>
      <title>The 32-bit Linux real and virtual memory boundaries</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060821c</link>

      <description>The x86 Linux kernel maps real memory into the
	address space of each process, and there is a performance
	advantage in the ability to map all of real memory in every
	process. On 32-bit CPUs the address space is limited to 4GiB,
	so one needs to decide how much to use to map real memory and
	how much for the use of each process.</description>
      <category>Linux</category>
      <category>kernel</category>
      <category>memory</category>
      <category>virtual memory</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060821c</guid>
      <pubDate>Mon, 21 Aug 2006 23:45:00 +0100</pubDate>
    </item>

    <item>
      <title>Another chip with 16 CPUs</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060821b</link>
      <description>Just seen another chip with 16 CPUs, this one
	not quite there yet but with the aim of having a grid
	style architecture.</description>
      <category>architecture</category>
      <category>hardware</category>
      <category>CPU</category>
      <category>MIPS</category>
      <category>Linux</category>
      <category>grid</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060821b</guid>
      <pubDate>Mon, 21 Aug 2006 13:35:00 +0100</pubDate>
    </item>

    <item>
      <title>15 years of Linux</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060821</link>
      <description>Linux was first announced 15 years ago. It is
	a long time, and perhaps Plan 9 or BSD should have
	enjoyed its popularity. But it is good enough and its
	licence is the GPL.</description>
      <category>Linux</category>
      <category>kernel</category>
      <category>operating system</category>
      <category>GPL</category>
      <category>BSD</category>
      <category>UNIX</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060821</guid>
      <pubDate>Mon, 21 Aug 2006 09:40:00 +0100</pubDate>
    </item>

    <item>
      <title>A chip with 16 CPUs</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060815</link>
      <description>Movidis is selling a server based on a low
	power 16-CPU chip by Cavium. Interesting tradeoffs,
	looks like (and I hope it is) a winner.</description>
      <category>architecture</category>
      <category>hardware</category>
      <category>CPU</category>
      <category>MIPS</category>
      <category>Linux</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060815</guid>
      <pubDate>Tue, 15 Aug 2006 18:15:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates to the font notes</title>
      <link>http://WWW.sabi.co.UK/Notes/linuxFonts.html#appBlog?060813</link>
      <description>Some updates to my fonts notes and the addition of a new
	section with pointers to font related entries in my blog.</description>
      <category>X</category>
      <category>Linux</category>
      <category>UNIX</category>
      <category>fonts</category>

      <pubDate>Sun, 13 Aug 2006 21:05:00 +0100</pubDate>
    </item>

    <item>
      <title>Finer tagging of text in HTML and the semantic web</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060810b</link>
      <description>My own conventions for tagging various parts of speech
	and types of discourse when I write HTML.</description>
      <category>XML</category>
      <category>HTML</category>
      <category>SGML</category>
      <category>markup</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060810b</guid>
      <pubDate>Thu, 10 Aug 2006 22:20:00 +0100</pubDate>
    </item>

    <item>
      <title> Fonts, antialiasing and low DPI</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060810</link>
      <description>While in general I reckon that font
	antialiasing is not a good idea and is unnecessary, on low
	DPI screen it can be the least bad of a bad lot. Also,
	some hints on how to scale up fonts on high DPI
	monitors.</description>
      <category>Linux</category>
      <category>fonts</category>
      <category>antialiasing</category>
      <category>X window system</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060810</guid>
      <pubDate>Thu, 10 Aug 2006 21:40:00 +0100</pubDate>
    </item>

    <item>
      <title>XML and HTML microformats</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060809</link>
      <description>Microformats are an objectionable way of
	describing data not with XML tags, but with HTML
	attributes. That is not quite nice. But much of the same
	effect can be obtained by describing text with HTML tags,
	and help data processing programs with
	attributes.</description>
      <category>XML</category>
      <category>HTML</category>
      <category>SGML</category>
      <category>markup</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060809</guid>
      <pubDate>Wed, 09 Aug 2006 19:05:00 +0100</pubDate>
    </item>

    <item>
      <title>What is the resident set size, 'exmap' and working sets</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060808</link>
      <description>
	The Linux resident set size of processes, a better way to
	compute it, and what would be a much better way to define it.
      </description>
      <category>Linux</category>
      <category>operating systems</category>
      <category>paging</category>
      <category>memory management</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060808</guid>
      <pubDate>Tue, 08 Aug 2006 20:55:00 +0100</pubDate>
    </item>

    <item>
      <title>Restructuring of the front page</title>
      <link>http://WWW.sabi.co.UK/?060807</link>
      <description>My site front page has been restructured (the
	content is almost the same) to use tables instead of
	lists, as that seems slightly more readable.</description>
      <category>web</category>
      <category>html</category>
      <category>design</category>

      <pubDate>Mon, 07 Aug 2006 15:20:00 +0100</pubDate>
    </item>

    <item>
      <title>Solved problem with automounted filesystems and 'updatedb'</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060806c</link>
      <description>The 'mlocate' version of 'updatedb' will not
	descend into an empty directory, and automounter mount
	points look like empty directories. The better solution is
	to change how 'updatedb' is used, to have a list of
	filesystems it should scan and a 'locate' database for
	each one (usually in each one).</description>
      <category>UNIX</category>
      <category>Linux</category>
      <category>filesystems</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060806c</guid>
      <pubDate>Sun, 06 Aug 2006 23:50:00 +0100</pubDate>
    </item>

    <item>
      <title>The 'init' daemon and 'inittab'</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060806b</link>
      <description>I am quite unhappy about the current
	structure of the GNU/Linux startup system, based on a
	misdesigned corruption of the original System V 'init'
	scheme. Which should be fixed, not corrupted, or replaced
	with something better.</description>
      <category>UNIX</category>
      <category>Linux</category>
      <category>pragmatics</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060806b</guid>
      <pubDate>Sun, 06 Aug 2006 23:10:00 +0100</pubDate>
    </item>

    <item>
      <title>An amusing example of (moderately) bad code</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060806</link>
      <description>
	Just spotted some (moderately) bad code in some kernel IDE
	driver. All too typical unfortunately.
      </description>
      <category>Linux</category>
      <category>software engineering</category>
      <category>pragmatics</category>
      <category>programming</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060806</guid>
      <pubDate>Sun, 06 Aug 2006 16:40:00 +0100</pubDate>
   </item>

    <item>
      <title>Using automounter maps instead of '/etc/fstab'</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060805</link>
      <description>
	Static mounting via '/etc/fstab' can be replaced with dynamic
	mounting via 'autofs' and there are some reasonable advantages.
      </description>
      <category>Linux</category>
      <category>filesystems</category>
      <category>partitions</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060805</guid>
      <pubDate>Sat, 05 Aug 2006 23:25:00 +0100</pubDate>
    </item>

    <item>
      <title>Small updates to my KDE startup script</title>
      <link>http://WWW.sabi.co.UK/#sourcesDokde?060805</link>
      <description>
	My alternative (to 'startkde') KDE startup is not updated for
	KDE 3.5.3, which has a slightly different startup sequence. It
	also contains a couple of fixes to bugs in 'startkde'.
      </description>
      <category>Linux</category>
      <category>KDE</category>
      <category>script</category>
      <category>shell</category>

      <pubDate>Sat, 05 Aug 2006 17:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Added review of the game "Sniper Elite"</title>
      <link>http://WWW.sabi.co.UK/Notes/miscGames.html#gameSniper?060729</link>
      <description>
	Reviewed the "Sniper Elite" game and slightly edited
	some of the other game reviews.
      </description>
      <category>videogames</category>
      <category>games</category>
      <category>reviews</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/miscGames.html#gameSniper?060729</guid>
      <pubDate>Sat, 29 Jul 2006 13:30:00 +0100</pubDate>
    </item>

    <item>
      <title>Using humans as information system components</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060728b</link>
      <description>
      </description>
      <category>architecture</category>
      <category>system design</category>
      <category>pattern matching</category>
      <category>web services</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060728b</guid>
      <pubDate>Fri, 28 Jul 2006 16:30:00 +0100</pubDate>
    </item>

    <item>
      <title>A bitter irony in the design of 'udev'</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060728</link>
      <description>
	The 'sysfs' file system defines files that for each
	device state which major and minor numbers its device
	files should have, but does not actually define the
	device file itself.
      </description>
      <category>Linux</category>
      <category>kernel</category>
      <category>udev</category>
      <category>devfs</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060728</guid>
      <pubDate>Fri, 28 Jul 2006 14:45:00 +0100</pubDate>
    </item>

    <item>
      <title>CPU bound modern games and top end GPUs</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060727b</link>
      <description>
	Comparing recent CPUs on recent games, these seem rather
	CPU bound.
      </description>
      <category>games</category>
      <category>videogames</category>
      <category>Athlon</category>
      <category>Core 2</category>
      <category>graphics</category>
      <category>CPU</category>
      <category>NVIDIA</category>
      <category>AMD</category>
      <category>Intel</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060727b</guid>
      <pubDate>Thu, 27 Jul 2006 23:25:00 +0100</pubDate>
    </item>

    <item>
      <title>Desktop GNU/Linux is better with 1GiB of RAM</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060727</link>
      <description>
	Another user has sold out and discovered that desktop GNU/Linux
	performs a lot better with 1GiB than with 512MiB of RAM.
      </description>
      <category>Linux</category>
      <category>RAM</category>
      <category>ECC</category>
      <category>1GiB</category>
      <category>swapping</category>
      <category>virtual memory</category>
      <category>KDE</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060727</guid>
      <pubDate>Thu, 27 Jul 2006 22:15:00 +0100</pubDate>
    </item>

    <item>
      <title>Another symptom of the Microsoft cultural hegemony</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060726</link>
      <description>
	Many people seem to think that in GNU/Linux command
	options should be last, just as in MS Windows.
      </description>
      <category>UNIX</category>
      <category>Linux</category>
      <category>MS Windows</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060726</guid>
      <pubDate>Wed, 26 Jul 2006 23:05:00 +0100</pubDate>
    </item>

    <item>
      <title>Interview with the CEO of NVIDIA</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060725b</link>
      <description>
	The CEO of NVIDIA speaks about programmability in
	graphics cards, and approves the Blu-Ray drive in the
	PS3, and the large investments companies have to put in
	development with a 6 month product generation cycle.
      </description>
      <category>PC</category>
      <category>games</category>
      <category>graphics</category>
      <category>NVIDIA</category>
      <category>PS/3</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060725b</guid>
      <pubDate>Tue, 25 Jul 2006 22:35:00 +0100</pubDate>
    </item>

    <item>
      <title>ATi and AMD merge</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060725</link>
      <description>
	Some comments on the AMD and ATi merger, which seems to
	me a bit strange, because the two companies are really
	not that related. I suspect that it is mainly a reaction
	to Intel's dominance of the integrated chipset market.
      </description>
      <category>PC</category>
      <category>games</category>
      <category>graphics</category>
      <category>VGA</category>
      <category>AMD</category>
      <category>ATi</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060725</guid>
      <pubDate>Tue, 25 Jul 2006 21:15:00 +0100</pubDate>
    </item>

    <item>
      <title>Mark Rein on games on laptops and Intel graphics chipsets</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060724d</link>
      <description>
	Mark Rein complains with some causes that most laptops
	and many desktops are sold with Intel graphics chips
	that cannot run the latest games, including those from
	his company. Too bad that game companies don't feel like
	selling to the majority of the market though. probably
	the issue is games with too much geometry, which is not
	as easily scalable as effects.
      </description>
      <category>PC</category>
      <category>games</category>
      <category>graphics</category>
      <category>VGA</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060724d</guid>
      <pubDate>Mon, 24 Jul 2006 22:50:00 +0100</pubDate>
    </item>

    <item>
      <title>Getting software into the Linux kernel</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060724c</link>
      <description>
	Controlling Linux kernel and base user level sources is
	precious, and some sources are merged in with a lot less
	fuss than others.
      </description>
      <category>Linux</category>
      <category>kernel</category>
      <category>udev</category>
      <category>Reiser4</category>
      <category>XFS</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060724c</guid>
      <pubDate>Mon, 24 Jul 2006 21:55:00 +0100</pubDate>
    </item>

    <item>
      <title>Partitions, extended partitions and 'ms-sys -p' on NTFS</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060724b</link>
      <description>
	The GNU/Linux utility 'ms-sys -p' fixes a number of
	partition issues for MS Windows, but it damages NTFS
	filesystems.
      </description>
      <category>Linux</category>
      <category>partitions</category>
      <category>MS Windows</category>
      <category>NTFS</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060724b</guid>
      <pubDate>Mon, 24 Jul 2006 21:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Very odd extra 16 bytes</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060724</link>
      <description>
	Mysteriously and temporarily a hard drive shifts its
	contents by 16 bytes.
      </description>
      <category>PC</category>
      <category>IO</category>
      <category>discs</category>
      <category>peripherals</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060724</guid>
      <pubDate>Mon, 24 Jul 2006 21:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Now IO has priorities too under Linux</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060723</link>
      <description>
	Just been told that the 'cfq' elevator under Linux can
	prioritize IO, which can help with interactive responsiveness.
      </description>
      <category>Linux</category>
      <category>IO</category>
      <category>latency</category>
      <category>performance</category>
      <category>games</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060723</guid>
      <pubDate>Sun, 23 Jul 2006 23:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Switched to SMP and voluntary preemption</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060713</link>
      <description>
	After a long time, my default Linux kernel build is now
	for SMP and voluntary preemption, even if I have a
	single CPU, just to improve interactive responsiveness.
	I have switched because the popularity of dual core
	chips means that the SMP and premption code is now an
	itch for more people, so probably it is more reliable
	than in the past.
      </description>
      <category>Linux</category>
      <category>SMP</category>
      <category>games</category>
      <category>latency</category>
      <category>preemption</category>
      <category>performance</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-3rd.html#060713</guid>
      <pubDate>Thu, 13 Jul 2006 20:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Fixed front/rear ALSA parametric devices</title>
      <link>http://WWW.sabi.co.UK/Cfg/ALSA/asound.conf?060704</link>
      <description>
	Oops, just realized that the definition of the
	front/rear split parametric PCMs was not quite right,
	and cleaned it up too.
      </description>
      <category>Linux</category>
      <category>sound</category>
      <category>ALSA</category>
      <category>asound.conf</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Cfg/ALSA/asound.conf?060704</guid>
      <pubDate>Tue, 04 Jul 2006 22:45:00 +0100</pubDate>
    </item>

    <item>
      <title>The 'ext4' file system and RHEL</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-3rd.html#060702</link>
      <description>
	There is now a plan to add 48 bit block addresses and
	extents to the 'ext3' file system, and that sounds odd
	as there are several other existing file systems with
	the same or better features. But RedHat has a large
	interest in extending 'ext3' rather than using a new
	file system type for RHEL.
      </description>
      <category>Linux</category>
      <category>file system</category>
      <category>ext3</category>
      <category>ext4</category>
      <category>JFS</category>
      <category>XFS</category>
      <category>Reiser4</category>
      <category>RedHat</category>
      <category>RHEL</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-3rd.html#060702</guid>
      <pubDate>Sun, 02 Jul 2006 23:00:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates list of file system description papers</title>
      <link>http://WWW.sabi.co.UK/Notes/linuxFS.html#fsRefsDesc?060702</link>
      <description>
	Added a couple of memos on the evolution of the 'ext3'
	file system into the new 'ext4' one.
      </description>
      <category>Linux</category>
      <category>file system</category>
      <category>ext3</category>
      <category>ext4</category>
      <category>RedHat</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/linuxFS.html#fsRefsDesc?060702</guid>
      <pubDate>Sun, 02 Jul 2006 22:55:00 +0100</pubDate>
    </item>

    <item>
      <title>Swap space misallocation in Linux</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060625b</link>
      <description>
	Linux allocates swap space blocks a bit too late, and
	this causes severe performance problems. There is a
	simple and general alternative.
      </description>
      <category>Linux</category>
      <category>virtual memory</category>
      <category>swapping</category>
      <category>paging</category>
      <category>file system</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060625b</guid>
      <pubDate>Sun, 25 Jun 2006 22:30:00 +0100</pubDate>
    </item>

    <item>
      <title>Partial specialization for APIs in C++</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060625</link>
      <description>
	If an interface depends on more than one implementation
	type C++ does not help a lot with avoiding code or API
	duplication, because of a specific restriction.
      </description>
      <category>C++</category>
      <category>template</category>
      <category>class</category>
      <category>interface</category>
      <category>implementation</category>
      <category>object oriented</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060625</guid>
      <pubDate>Sun, 25 Jun 2006 21:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Sausages: 93% pork, but perhaps 55% pork meat</title>
      <link>http://WWW.sabi.co.UK/Notes/miscShopping#msBritishSausages</link>
      <description>
	A review of MS "british premium pork sausages", whose
	ingredients include 93% "pork", but have perhaps 50% content of
	apparently no nutritional value, and the protein content that
	may imply perhaps only 55% pork meat overall.
      </description>
      <category>shopping</category>
      <category>pork</category>
      <category>sausage</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/miscShopping#msBritishSausages</guid>
      <pubDate>Thu, 15 Jun 2006 23:25:00 +0100</pubDate>
    </item>

    <item>
      <title>Sustainable ADSL traffic</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060612</link>
      <description>
	The results of some tests on the sustainable uplink and
	downlink bandwidths for my ADSL line.
      </description>
      <category>ADSL</category>
      <category>bandwidth</category>
      <category>networking</category>
      <category>traffic shaping</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060612</guid>
      <pubDate>Mon, 12 Jun 2006 20:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Small updates to my IP shaping script</title>
      <link>http://WWW.sabi.co.UK/#sourcesSabishape?060608</link>
      <description>
	I have fixed a couple of small improprieties; also
	introduced a distinction in the traffic classes between
	the ceiling bandwidth (same as total for all classes)
	and target rate, whose sum total is now lower, not
	higher than the total for their parent.
      </description>
      <category>Linux</category>
      <category>networking</category>
      <category>Netfilter</category>
      <category>iptables</category>
      <category>shell</category>
      <category>shaping</category>
      <category>IP</category>
      <category>configuration</category>
      <category>traffic control</category>

      <pubDate>Thu, 08 Jun 2006 18:30:00 +0100</pubDate>
    </item>

    <item>
      <title>Constructor, destructors, C++ and terminology</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060606b</link>
      <description>
	What ''constructors'' should do in C++, what the terms
	''object'', ''constructor'' and 'destructor'' mean in
	C++ and in computer science, and what ''destructors''
	should do in C++.
      </description>
      <category>C++</category>
      <category>object</category>
      <category>constructor</category>
      <category>destructor</category>
      <category>class instance</category>
      <category>object oriented</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060606b</guid>
      <pubDate>Tue, 06 Jun 2006 23:00:00 +0100</pubDate>
    </item>

    <item>
      <title>The new Microsoft ClearType oriented fonts</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060606</link>
      <description>
	Found some screenshots of the new Microsoft Vista fonts
	which are ClearType oriented. They are nice, but the
	fringing is bad and the hinting makes them look bad
	without antialising.
      </description>
      <category>Microsoft</category>
      <category>fonts</category>
      <category>ClearType</category>
      <category>antialiasing</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060606</guid>
      <pubDate>Tue, 06 Jun 2006 18:35:00 +0100</pubDate>
    </item>

    <item>
      <title>No to keyboards with black keys</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060604</link>
      <description>
	In review of 10 keyboards 9 had black keys with white
	lettering. It looks cool but with time it becomes unreadable.
      </description>
      <category>hardware</category>
      <category>keyboard</category>
      <category>ergonomics</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060604</guid>
      <pubDate>Sun, 04 Jun 2006 16:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Linux based PS3 as PC replacement</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060603</link>
      <description>
	I had speculated that Sony sees a PS/3 with Linux as a
	PC replacement, not just a games console. Sony's head of
	development has confirmed this quite clearly.
      </description>
      <category>PS/3</category>
      <category>Sony</category>
      <category>Microsoft</category>
      <category>marketing</category>
      <category>videogames</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060603</guid>
      <pubDate>Sat, 03 Jun 2006 10:05:00 +0100</pubDate>
    </item>

    <item>
      <title>C++ and separating interface from implementation</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060602</link>
      <description>
	How to separate interface from implementation in C++, in
	particular when the data representation is different,
	but the API is equivalent.
      </description>
      <category>C++</category>
      <category>templates</category>
      <category>inheritance</category>
      <category>interface</category>
      <category>implementation</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060602</guid>
      <pubDate>Fri, 02 Jun 2006 23:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates to file system notes</title>
      <link>http://WWW.sabi.co.UK/Notes/linuxFS.html?060602</link>
      <description>
	Many updates to my page of notes on Linux file systems,
	mostly copied from the tech blog with some updates.
      </description>
      <category>Linux</category>
      <category>file system</category>
      <category>ext3</category>
      <category>JFS</category>
      <category>XFS</category>
      <category>ReiserFS</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/linuxFS.html?060602</guid>
      <pubDate>Fri, 02 Jun 2006 19:50:00 +0100</pubDate>
    </item>

    <item>
      <title>Coarse parallelism and pipeline duplication</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060528</link>
      <description>
	Some more illustration of what coarse parallelism in
	games could be like, for example for characters.
      </description>
      <category>videogames</category>
      <category>Sony</category>
      <category>PS3</category>
      <category>parallelism</category>
      <category>Cell</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060528</guid>
      <pubDate>Sun, 28 May 2006 22:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Non-virtual destructors in C++</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060524</link>
      <description>
	During a discussion of non-virtual destructors in C++,
	someone suggested that they can be used to indicate a
	class should not be derived for. Possibly, but it goes
	against the grain...
      </description>
      <category>C++</category>
      <category>virtual</category>
      <category>Java</category>
      <category>final</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060524</guid>
      <pubDate>Wed, 24 May 2006 23:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates to KDE notes</title>
      <link>http://WWW.sabi.co.UK/Notes/linuxKDE.html?060515</link>
      <description>
	Some revision and extension of my KDE notes, in particular
	added details on how to do some particular tasks.
      </description>
      <category>KDE</category>
      <category>hints</category>
      <category>X</category>
      <category>Linux</category>
      <category>performance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/linuxKDE.html?060515</guid>
      <pubDate>Mon, 15 May 2006 18:00:00 +0100</pubDate>
    </item>

    <item>
      <title>Quick write speed test for CIFS and NFS</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060514</link>
      <description>
	A quick write speed test of CIFS and NFS, Linux-to-Linux
	over a 100MHz network. Unexpectedly writing lots of
	small files is slower with CIFS than NFS in asynchronous
	mode. Either CIFS does not run in asynchronous mode or
	perhaps lots more system calls is the problem.
      </description>
      <category>Linux</category>
      <category>file system</category>
      <category>SMB</category>
      <category>NFS</category>
      <category>CIFS</category>
      <category>Samba</category>
      <category>Ethernet</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060514</guid>
      <pubDate>Sun, 14 May 2006 19:25:00 +0100</pubDate>
    </item>

    <item>
      <title>Quick read speed test for CIFS and NFS</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060513</link>
      <description>
	A quick read speed test of CIFS and NFS, Linux-to-Linux
	over a 100MHz network. As expected usually CIFS is preferable.
      </description>
      <category>Linux</category>
      <category>file system</category>
      <category>SMB</category>
      <category>NFS</category>
      <category>CIFS</category>
      <category>Samba</category>
      <category>Ethernet</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060513</guid>
      <pubDate>Sat, 13 May 2006 20:15:00 +0100</pubDate>
    </item>

    <item>
      <title>What is OO and why it works</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060512c</link>
      <description>
	A reminder of what OO is about and why it works. it is a
	program decomposition paradigm where each 
      </description>
      <category>programming</category>
      <category>software engineering</category>
      <category>object oriented</category>
      <category>modules</category>
      <category>overloading</category>
      <category>genericity</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060512c</guid>
      <pubDate>Fri, 12 May 2006 20:50:00 +0100</pubDate>
    </item>

    <item>
      <title>Inlining at the call site</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060512b</link>
      <description>
	Many important optimisations like inlining should apply
	not to the definition of functions or classes, but to
	their instantiations.
      </description>
      <category>programming</category>
      <category>software engineering</category>
      <category>optimization</category>
      <category>inlining</category>
      <category>generics</category>
      <category>overloading</category>
      <category>templates</category>
      <category>inheritance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060512b</guid>
      <pubDate>Fri, 12 May 2006 19:00:00 +0100</pubDate>
    </item>

    <item>
      <title>Open, close or manifest type systems, programs and libraries</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060512</link>
      <description>
	Type systems can be open, closed or manifest, and code
	can belong to programs or libraries.
      </description>
      <category>programming</category>
      <category>type systems</category>
      <category>software engineering</category>
      <category>optimization</category>
      <category>C</category>
      <category>C++</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060512</guid>
      <pubDate>Fri, 12 May 2006 18:20:00 +0100</pubDate>
    </item>

    <item>
      <title>PS3 Gran Turismo HD halfway demo</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060511</link>
      <description>
	Some comments on the importance of Gran Turismo as a
	Playstation signature game, and coarse parallelism,
	relating to reports of an incomplete demo of the PS3
	demo for the game.
      </description>
      <category>videogames</category>
      <category>Sony</category>
      <category>PS3</category>
      <category>parallelism</category>
      <category>Cell</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060511</guid>
      <pubDate>Thu, 11 May 2006 09:45:00 +0100</pubDate>
    </item>

    <item>
      <title>Quick speed test for Reiser4</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060510</link>
      <description>
	A quick speed test of Reiser4, with JFS as a reference
	point, on my root filesystem. Looks good, but seems to
	delay writing out modified inodes for a long time.
      </description>
      <category>Linux</category>
      <category>file system</category>
      <category>JFS</category>
      <category>Reiser4</category>
      <category>performance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060510</guid>
      <pubDate>Wed, 10 May 2006 13:55:00 +0100</pubDate>
    </item>

    <item>
      <title>Added review of Primark basic clothes</title>
      <link>http://WWW.sabi.co.UK/Notes/miscShopping.html#primarkClothes?060509</link>
      <description>
	Primark sells basic clothes that are cheap and of
	reasonable quality.
      </description>
      <category>Shopping</category>
      <category>clothes</category>
      <category>polo shirts</category>
      <category>underwear</category>
      <category>Primark</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/miscShopping.html#primarkClothes?060509</guid>
      <pubDate>Tue, 09 May 2006 17:45:00 +0100</pubDate>
    </item>

    <item>
      <title>Summarizing the font situation</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060509</link>
      <description>
	A simpler summary of what to do to have nice fonts with
	GNU/Linux, at least in X.
      </description>
      <category>Linux</category>
      <category>X11</category>
      <category>fonts</category>
      <category>FreeType</category>
      <category>Type 1</category>
      <category>TrueType</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-2nd.html#060509</guid>
      <pubDate>Tue, 09 May 2006 16:50:00 +0100</pubDate>
    </item>

    <item>
      <title>A nice working IT8212F chipset card</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060507</link>
      <description>
	Accidentally obtained a second IT8212F based ATA host
	adapter, and this one actually works fairly well.
      </description>
      <category>Linux</category>
      <category>kernel</category>
      <category>driver</category>
      <category>ATA</category>
      <category>SCSI</category>
      <category>RAID</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-2nd.html#060507</guid>
      <pubDate>Sun, 05 May 2006 19:20:00 +0100</pubDate>
    </item>

    <item>
      <title>Using SSL with name based websites</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060505b</link>
      <description>
	Usually it is believed that there only be one SSL
	enabled web site per IP address, but there are at least
	two important and useful special cases where there can
	be multiple ones.
      </description>
      <category>encryption</category>
      <category>SSL</category>
      <category>HTTP</category>
      <category>web</category>
      <category>Apache</category>
      <category>SSL</category>
      <category>OpenSSL</category>
      <category>web server</category>
      <category>virtual hosts</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060505b</guid>
      <pubDate>Fri, 05 May 2006 22:50:00 +0100</pubDate>
    </item>

    <item>
      <title>Two papers on costs of garbage collection and reference counts</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060505</link>
      <description>
	Comments on two interesting papers on the space costs of
	delayed reference counting, and the time cost of tracing
	and delayed deallocation in garbage collection.
      </description>
      <category>garbage collection</category>
      <category>reference counting</category>
      <category>memory</category>
      <category>memory management</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060505</guid>
      <pubDate>Fri, 05 May 2006 21:05:00 +0100</pubDate>
    </item>

    <item>
      <title>Zalman Fanmate2 review</title>
      <link>http://WWW.sabi.co.UK/Notes/hwNotesReviews.html#zalmanFanmate2?060505</link>
      <description>
	A positive review of the Zalman Fanmate2 fan RPM
	adjuster. It does not become hot.
      </description>
      <category>hardware</category>
      <category>cooling</category>
      <category>noise</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/hwNotesReviews.html#zalmanFanmate2?060505</guid>
      <pubDate>Fri, 05 May 2006 16:35:00 +0100</pubDate>
    </item>

    <item>
      <title>Turning down fan RPMs for a quieter PC</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060504</link>
      <description>
	Thanks to a convenient Zalman Fanmate2 I have reduced
	the speed of both my PCU and rear case fans to the point
	I can't hear them anymore, and without much if any
	increase in temperatures.
      </description>
      <category>hardware</category>
      <category>cooling</category>
      <category>noise</category>
      <category>Athlon</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060504</guid>
      <pubDate>Thu, 04 May 2006 20:20:00 +0100</pubDate>
    </item>

    <item>
      <title>Halo for PC and what is a CPU bound games</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060502c</link>
      <description>
	Halo for is a lot faster on my new CPU than the slower
	one. Clearly CPU bound. Was this intentional?
      </description>
      <category>videogames</category>
      <category>performance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060502c</guid>
      <pubDate>Tue, 02 May 2006 23:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates to the ALSA sound notes and configuration</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060502b</link>
      <description>
	Updates the the Linux ALSA sound notes and sample ALSA
	lib configurations.
      </description>
      <category>Linux</category>
      <category>sound</category>
      <category>ALSA</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060502b</guid>
      <pubDate>Tue, 02 May 2006 18:20:00 +0100</pubDate>
    </item>

    <item>
      <title>Discoveries about FontConfig configuration files</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060502</link>
      <description>
	
      </description>
      <category>Linux</category>
      <category>X11</category>
      <category>fonts</category>
      <category>FontConfig</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-2nd.html#060502</guid>
      <pubDate>Tue, 02 May 2006 15:50:00 +0100</pubDate>
    </item>

    <item>
      <title>Sample FontConfig configuration files</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060430</link>
      <description>
	
      </description>
      <category>Linux</category>
      <category>X11</category>
      <category>fonts</category>
      <category>FontConfig</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Notes/anno06-2nd.html#060430</guid>
      <pubDate>Sun, 30 Apr 2006 22:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Fixed ALSA parametric devices</title>
      <link>http://WWW.sabi.co.UK/Cfg/ALSA/asound.conf?060430</link>
      <description>
	There was a bizarre but fatal problem in my parametric
	'asound.conf' in the 'pcm.rec'/'dsnoop' configuration.
	It is fixed now. Also, small updates and cleanups.
      </description>
      <category>Linux</category>
      <category>sound</category>
      <category>ALSA</category>
      <category>asound.conf</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Cfg/ALSA/asound.conf?060430</guid>
      <pubDate>Sun, 30 Apr 2006 22:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates to X11 sample config files</title>
      <link>http://WWW.sabi.co.UK/Cfg/X11/?060430</link>
      <description>
	Added DejaVu aliases.
      </description>
      <category>Linux</category>
      <category>X11</category>
      <category>fonts</category>

      <guid isPermaLink="true"
	>http://WWW.sabi.co.UK/Cfg/X11/?060420</guid>
      <pubDate>Sun, 30 Apr 2006 21:30:00 +0100</pubDate>
    </item>

    <item>
      <title>Freetype2 font rendering and quality</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060429</link>
      <description>
	The Linux fonts mess: what are the best options for
	rasterization quality? Options and comparisons.
      </description>
      <category>Linux</category>
      <category>fonts</category>
      <category>TrueType</category>
      <category>Type1</category>
      <category>X11</category>
      <category>FontConfig</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060429</guid>
      <pubDate>Sat, 29 Apr 2006 20:05:00 +0100</pubDate>
    </item>

    <item>
      <title>When did the UNIX style start to fade?</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060424c</link>
      <description>
	The UNIX style of pragmatics is ever more forgotten...
	When did the fade begin? I suspect that was with
	SystemIII and 'fsck'.
      </description>
      <category>Linux</category>
      <category>UNIX</category>
      <category>pragmatics</category>
      <category>style</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060424c</guid>
      <pubDate>Mon, 24 Apr 2006 23:50:00 +0100</pubDate>
    </item>

    <item>
      <title>Summary of fsck times</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060424b</link>
      <description>
	Gathering some 'fsck' times for two rather different
	filesystems, and observations: time mostly proportional
	to number of inodes, can be pretty long for large
	filesystems, even in the best case.
      </description>
      <category>Linux</category>
      <category>file system</category>
      <category>ext3</category>
      <category>JFS</category>
      <category>XFS</category>
      <category>ReiserFS</category>
      <category>fsck</category>
      <category>performance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060424b</guid>
      <pubDate>Mon, 24 Apr 2006 23:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Some larger filesystem informal speed tests</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060424</link>
      <description>
	Some speed tests on a filesystem with 65GiB and 85k
	files in it, under 'ext3', JFS, XFS, ReiserFS.
      </description>
      <category>Linux</category>
      <category>ext3</category>
      <category>file system</category>
      <category>JFS</category>
      <category>XFS</category>
      <category>ReiserFS</category>
      <category>performance</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060424</guid>
      <pubDate>Mon, 24 Apr 2006 23:10:00 +0100</pubDate>
    </item>

    <item>
      <title>Distasteful Microsoft bashing</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060423</link>
      <description>
	I have been sad to read a series of quotes bashing
	Microsoft when they both are instead healthy self
	criticism and they bash technical and project management
	issues that are problems with free software too. Free
	software is about freedom.
      </description>
      <category>GNU</category>
      <category>microsoft</category>
      <category>free software</category>
      <category>opensource</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060423</guid>
      <pubDate>Sun, 23 Apr 2006 19:55:00 +0100</pubDate>
    </item>

    <item>
      <title>Disc-to-disc defragmenting and backups</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060422b</link>
      <description>
	Defragmentation is often best done by disc-to-disc backup
	(image) and restore (tree). But then backup should be
	done that way too.
      </description>
      <category>Linux</category>
      <category>backup</category>
      <category>performance</category>
      <category>fragmentation</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060422b</guid>
      <pubDate>Sat, 22 Apr 2006 23:50:00 +0100</pubDate>
    </item>

    <item>
      <title>Another file system speed test</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060422</link>
      <description>
	Some comments on what seems to me a rather flawed and
	incomplete speed test of a few file systems. Still a lot
	better than the previous one I had seen.
      </description>
      <category>Linux</category>
      <category>file system</category>
      <category>JFS</category>
      <category>XFS</category>
      <category>ReiserFS</category>
      <category>ext3</category>
      <category>performance</category>
      <category>fragmentation</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060422</guid>
      <pubDate>Sat, 22 Apr 2006 13:15:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates to Linux font notes</title>
      <link>http://WWW.sabi.co.UK/Notes/linuxFonts.html?060420</link>
      <description>
	Updates to my Linux font notes, with some extensions.
      </description>
      <category>Linux</category>
      <category>fonts</category>
      <category>TrueType</category>
      <category>Type1</category>
      <category>X windows</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/linuxFonts.html?060420</guid>
      <pubDate>Thu, 20 Apr 2006 21:20:00 +0100</pubDate>
    </item>

    <item>
      <title>Updates to Linux coding pragmatics</title>
      <link>http://WWW.sabi.co.UK/Notes/linuxPragmaCoding.html#pathsDir</link>
      <description>
	Added to the Linux coding pragmatics recommendations to
	open files, especially configuration files, following a
	directory path, and what the default path should look like.
      </description>
      <category>UNIX</category>
      <category>Linux</category>
      <category>programming</category>
      <category>pragmatics</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/hwNotesReviews.html#pathsDir</guid>
      <pubDate>Thu, 20 Apr 2006 14:00:00 +0100</pubDate>
    </item>

    <item>
      <title>Online non-MMORPG game usage statistics</title>

      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060419</link>
      <description>
	GameSpy have a page with usage statistics for many
	popular non-MMORPG online games. Well, they aren't so
	popular compared to MMORPGs, except for CounterStrike.
      </description>
      <category>games</category>
      <category>MMORPG</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060419</guid>
      <pubDate>Wed, 19 Apr 2006 23:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Some speed/compatibility tests of USB2/FW chipsets</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060418</link>
      <description>
	I have recently bought a new USB2/FW PCI card, and
	somewhat earlier a new 3.5" external USB2/FW enclosure,
	so I now have two of each; so i tested the various
	combinations for compatibility and speed. The results
	are as usual a bit depressing.
      </description>
      <category>Linux</category>
      <category>hardware</category>
      <category>USB2</category>
      <category>IEEE1394</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060418</guid>
      <pubDate>Tue, 18 Apr 2006 21:15:00 +0100</pubDate>
    </item>

    <item>
      <title>No Name ULT31311 review</title>
      <link>http://WWW.sabi.co.UK/Notes/hwNotesReviews.html#nonameULT31311</link>
      <description>
	Review of an ULT31311 external USB2/FW box for 3.5"
	drives. It has some significant issues, and overall I can
	tolerate it, but I wish I had bought another one.
      </description>
      <category>hardware</category>
      <category>USB2</category>
      <category>IEEE1394</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/hwNotesReviews.html#nonameULT31311</guid>
      <pubDate>Tue, 18 Apr 2006 20:05:00 +0100</pubDate>
    </item>

    <item>
      <title>No Name TT-346U2F review</title>
      <link>http://WWW.sabi.co.UK/Notes/hwNotesReviews.html#nonameTT346U2F</link>
      <description>
	Review of a TT-346U2F external USB2/FW box for 5.25" and
	3.5" drives. It has some issues, but I like it.
      </description>
      <category>hardware</category>
      <category>USB2</category>
      <category>IEEE1394</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/hwNotesReviews.html#nonameTT346U2F</guid>
      <pubDate>Tue, 18 Apr 2006 19:15:00 +0100</pubDate>
    </item>

    <item>
      <title>Solid, HAL, D-BUS, sysfs</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060416b</link>
      <description>
	Solid is the new KDE hardware abstraction layer, and is
	build on top of various other already hairy layers. Not
	the UNIX way, more like the Microsoft/Debian way.
      </description>
      <category>Linux</category>
      <category>hardware</category>
      <category>autoconfiguration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060416b</guid>
      <pubDate>Sun, 16 Apr 2006 20:40:00 +0100</pubDate>
    </item>

    <item>
      <title>Retesting JFS performance over time</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-2nd.html#060416</link>
      <description>
	More testing of JFS performance drop over time as
	files get rewritten, this time after a full upgrade from
	Fedora 4 to Fedora 5.
      </description>
      <category>JFS</category>
      <category>file system</category>
      <category>Linux</category>
      <category>performance</category>
      <category>fragmentation</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-2nd.html#060416</guid>
      <pubDate>Sun, 16 Apr 2006 14:15:00 +0100</pubDate>
    </item>

    <item>
      <title>Idiotic automagic in Fedora 5</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060327</link>
      <description>
	The script that start 'xfs' in Fedora 5 overwrites font
	configuration files, and wrongly too.
      </description>
      <category>Fedora</category>
      <category>autoconfiguration</category>
      <category>fonts</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060327</guid>
      <pubDate>Thu, 27 Mar 2006 20:15:00 +0000</pubDate>
    </item>

    <item>
      <title>Impressive IO speedup on upgrading the CPU</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060326</link>
      <description>
	Upgrading my CPU and motherboard to significantly
	faster ones results in massive improvement to the speed
	of reading from hard drives under Linux, it looks like
	IO under Linux is CPU bound, amazingly.
      </description>
      <category>Linux</category>
      <category>IO</category>
      <category>Athlon 64</category>
      <category>speed</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060326</guid>
      <pubDate>Wed, 26 Mar 2006 22:05:00 +0000</pubDate>
    </item>

    <item>
      <title>Upgrading to Fedora 5</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060325</link>
      <description>
	After several months of using Fedora 4 I have upgraded
	to Fedora 5. Overall I am satisfied, even if I have
	disabled some of the more deleterious hacks.
      </description>
      <category>Linux</category>
      <category>Fedora</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060325</guid>
      <pubDate>Tue, 25 Mar 2006 21:55:00 +0000</pubDate>
    </item>

    <item>
      <title>Linux written page flushing can be fixed</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060323</link>
      <description>
	Just discovered that someone has found why modified file
	pages can take a long time to be written back to disc
	under Linux, irrespetive of the relevant kernel
	parameters, and has produced a very useful patch.
      </description>
      <category>Linux</category>
      <category>IO</category>
      <category>buffering</category>
      <category>hard disc</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060323</guid>
      <pubDate>Thu, 23 Mar 2006 19:20:00 +0000</pubDate>
    </item>

    <item>
      <title>Test of recent PC games: CPU or GPU bound?</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060322</link>
      <description>
	A test of recent PC games as to when they become CPU
	bound. The outcome is that a fast CPU is a good idea...
      </description>
      <category>videogames</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060322</guid>
      <pubDate>Wed, 22 Mar 2006 23:50:00 +0000</pubDate>
    </item>

    <item>
      <title>FT article on Godfather videogame</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060320</link>
      <description>
	The Financial Times writes about the impending release
	of the Godfather videogame, whose delayed release cost
	EA US$800m in its market value.
      </description>
      <category>videogames</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060320</guid>
      <pubDate>Mon, 20 Mar 2006 16:30:00 +0000</pubDate>
    </item>

    <item>
      <title>Small addition to Netfilter shell script</title>
      <link>http://WWW.sabi.co.UK/#sourcesSabifire?060320</link>
      <description>
	To make it easier for new users to understand the
	firewall building script I have added two new options
	to print 'iptables' commands before they are executed
	and to not execute them.
      </description>
      <category>Linux</category>
      <category>networking</category>
      <category>Netfilter</category>
      <category>iptables</category>
      <category>shell</category>
      <category>firewall</category>
      <category>IP</category>
      <category>IPv6</category>
      <category>configuration</category>

      <pubDate>Mon, 20 Mar 2006 16:05:00 +0000</pubDate>
    </item>

    <item>
      <title>Added a firewall configuration script, reflections</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060319c</link>
      <description>
	After adding a shell script for firewall configuration,
	reflections on how subtle and complex it is, and the
	firewall configuration it generates too, and how users
	mistakes physical accessibility of the sources and
	distributions of free software with their accessibility
	in a skills sense.
      </description>
      <category>Linux</category>
      <category>networking</category>
      <category>Netfilter</category>
      <category>iptables</category>
      <category>shell</category>
      <category>firewall</category>
      <category>IP</category>
      <category>IPv6</category>
      <category>configuration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060319c</guid>
      <pubDate>Sun, 19 Mar 2006 23:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Shell script for creating IP/IPv6 Netfilter firewalls</title>
      <link>http://WWW.sabi.co.UK/#sourcesSabifire</link>
      <description>
	Just added a shell script to set up a leaf node firewall
	configuration for Linux using Netfilter.
      </description>
      <category>Linux</category>
      <category>networking</category>
      <category>Netfilter</category>
      <category>iptables</category>
      <category>shell</category>
      <category>firewall</category>
      <category>IP</category>
      <category>IPv6</category>
      <category>configuration</category>

      <pubDate>Sun, 19 Mar 2006 23:45:00 +0000</pubDate>
    </item>

    <item>
      <title>The ClearSpeed architecture may interest AMD</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060319b</link>
      <description>
	AMD may be interested in using a ClearSpeed coprocessor
	to enhance the floating point performance of its CPUs.
	The ClearSpeed architecture presents some challenges.
      </description>
      <category>CPU</category>
      <category>parallelism</category>
      <category>AMD</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060319b</guid>
      <pubDate>Sun, 19 Mar 2006 22:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Cleaned the case filter, temperatures drop</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060319</link>
      <description>
	After cleaning the lower front filter for the air intake
	of my PC case temperatures of the hard disks and the CPU
	and chipset dropped significantly.
      </description>
      <category>hardware</category>
      <category>cooling</category>
      <category>case</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060319</guid>
      <pubDate>Sun, 19 Mar 2006 22:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Finally figured out ALSA parametric devices</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060315b</link>
      <description>
	The syntax for defining and making use of ALSA parametric
	devices seems obscure, but I have eventually managed to
	figure it out, at least in part.
      </description>
      <category>Linux</category>
      <category>sound</category>
      <category>ALSA</category>
      <category>asound.conf</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060315b</guid>
      <pubDate>Wed, 15 Mar 2006 23:55:00 +0000</pubDate>
    </item>

    <item>
      <title>Usual not very useful error messages</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060315</link>
      <description>
	Classic example of not very useful messages because of
	their incompleteness.
      </description>
      <category>Linux</category>
      <category>sound</category>
      <category>ALSA</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060315</guid>
      <pubDate>Wed, 15 Mar 2006 19:20:00 +0000</pubDate>
    </item>

    <item>
      <title>19" screens smaller than 17" ones, diagonals, ratios</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060314c</link>
      <description>
	A 19" inch widescreen monitor has less pixels than a 17"
	one. Quoting monitor diagonals only makes much sense
	if the aspect ratio is not much more oblong than the
	classic 4x3 one.
      </description>
      <category>hardware</category>
      <category>LCD</category>
      <category>monitor</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060314c</guid>
      <pubDate>Tue, 14 Mar 2006 18:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Sort order, internet domain names, IP addresses</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060314b</link>
      <description>
	An important property in UNIX like systems is that data
	be easily sortable. It is very annoying that internet
	domain names and IP addresses in dotted quad notation
	don't sort naturally.
      </description>
      <category>UNIX</category>
      <category>pragmatics</category>
      <category>programming</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060314b</guid>
      <pubDate>Tue, 14 Mar 2006 17:50:00 +0000</pubDate>
    </item>

    <item>
      <title>256MiB vs. 512MiB for top end videocards</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060314</link>
      <description>
	Nice test of two otherwise very similar top end
	videocards, with 256MiB and 512MiB. At top quality
	several modern games use way more than 256MiB of texture
	memory. Indeed at lower resolutions and memory they use
	more than 128MiB...
      </description>
      <category>videocards</category>
      <category>videogames</category>
      <category>NVIDIA</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060314</guid>
      <pubDate>Tue, 14 Mar 2006 17:35:00 +0000</pubDate>
    </item>

    <item>
      <title>Measuring the CPU cost of shared libraries</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060313</link>
      <description>
	Someone has checked how much shared libraries cost in
	CPU time. Not a lot in itself, but it adds up to a
	significant overhead.
      </description>
      <category>Linux</category>
      <category>shared libraries</category>
      <category>ELF</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060313</guid>
      <pubDate>Mon, 13 Mar 2006 20:20:00 +0000</pubDate>
    </item>

    <item>
      <title>Ubuntu, the enterprise level distribution</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060312d</link>
      <description>
	Mark Shuttleworth proposes a delay in the next Ubuntu
	release, and his reasons have to do with polishing it
	into an enterprise level distribution
      </description>
      <category>Linux</category>
      <category>Debian</category>
      <category>Ubuntu</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060312d</guid>
      <pubDate>Sun, 12 Mar 2006 15:55:00 +0000</pubDate>
    </item>

    <item>
      <title>Cell used in IBM blade servers</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060312c</link>
      <description>
	Using the Cell processor in IBM blade server, some not
	so fresh news.
      </description>
      <category>Cell</category>
      <category>PS/3</category>
      <category>IBM</category>
      <category>Sony</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060312c</guid>
      <pubDate>Sun, 12 Mar 2006 15:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Proof-of-concept virtual machine rootkits</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060312b</link>
      <description>
	Microsoft and UofM have crated some virtual machine
	based rootkits as a proof of concept. Isn't DRM based
	on VM technology similar?
      </description>
      <category>Microsoft</category>
      <category>security</category>
      <category>virtual machine</category>
      <category>DRM</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060312b</guid>
      <pubDate>Sun, 12 Mar 2006 15:35:00 +0000</pubDate>
    </item>

    <item>
      <title>More on launch and pricing of the PS/3</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060312</link>
      <description>
	Comments on Sony's plans to deliver PS/3 before the end
	of the year, on the meaning of 'spring' as a delivery
	date, and Sony's hopes that in Europe those that
	purchase Xbox 360 will also buy a PS/3, despite the high
	prices of both.
      </description>
      <category>Xbox 360</category>
      <category>PS/3</category>
      <category>Dell</category>
      <category>Sony</category>
      <category>Nintendo</category>
      <category>Microsoft</category>
      <category>videogames</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060312</guid>
      <pubDate>Sun, 12 Mar 2006 14:30:00 +0000</pubDate>
    </item>

    <item>
      <title>Launch and pricing for PS/3, overlap with PC price range</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060309b</link>
      <description>
	Sony is sticking to its promise of a spring 2006 launch
	for PS/3, and there is speculation on the launch price.
	Probably PS/3 will be launched at the end of June 2006,
	in very limited quantities, just the like Xbox 360, and
	at a price, like the Xbox 360, which overlaps with the
	price range of low end PCs.
      </description>
      <category>Xbox 360</category>
      <category>PS/3</category>
      <category>Dell</category>
      <category>Sony</category>
      <category>Nintendo</category>
      <category>Microsoft</category>
      <category>videogames</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060309b</guid>
      <pubDate>Thu, 09 Mar 2006 18:55:00 +0000</pubDate>
    </item>

    <item>
      <title>The Nintendo Revolution is shrewdly designed</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060309</link>
      <description>
	The specification for the Nintendo Revolution are not
	yet public, but it seems like that it will have fewer
	processors and much more cache than PS/3 or Xbox 360,
	and this will make it rather competitive.
      </description>
      <category>Revolution</category>
      <category>PS/3</category>
      <category>Xbox 360</category>
      <category>Sony</category>
      <category>Nintendo</category>
      <category>Microsoft</category>
      <category>videogames</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060309</guid>
      <pubDate>Thu, 09 Mar 2006 17:35:00 +0000</pubDate>
    </item>

    <item>
      <title>Retesting JFS performance over time</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060306b</link>
      <description>
	After two months of using my new fast 250GB discs I am
	retesting how much performance in my root partition has
	degraded. About twice, which is not too bad.
      </description>
      <category>JFS</category>
      <category>file system</category>
      <category>Linux</category>
      <category>performance</category>
      <category>fragmentation</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060306b</guid>
      <pubDate>Mon, 06 Mar 2006 20:05:00 +0000</pubDate>
    </item>

    <item>
      <title>Intel is leaving HyperThreading behind</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060306</link>
      <description>
	Intel will stop doing hyperthreaded CPUs, as it is
	switching to multiple cores. But recent game consoles
	are still in effect HyperThreaded. Hard to exploit.
      </description>
      <category>Intel</category>
      <category>HyperThreading</category>
      <category>dual core</category>
      <category>games</category>
      <category>PS/3</category>
      <category>Xbox/360</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060306</guid>
      <pubDate>Mon, 06 Mar 2006 19:45:00 +0000</pubDate>
    </item>

    <item>
      <title>Next generation game team sizes, one data point</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060305</link>
      <description>
	Lionhead are downsizing, and give numbers on how many
	developers they retain for how many projects, giving
	a rough indication of next gen team size.
      </description>
      <category>videogames</category>
      <category>PS/3</category>
      <category>Xbox/360</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060305</guid>
      <pubDate>Sun, 05 Mar 2006 14:15:00 +0000</pubDate>
    </item>

    <item>
      <title>Source released for EECH, active community</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060226c</link>
      <description>
	Source has been released a while ago for the splendid
	EECH games (Comanche Hokum and Apache Havoc), and a
	still active community produces extension and mods.
      </description>
      <category>videogames</category>
      <category>simulation</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060226c</guid>
      <pubDate>Sun, 26 Feb 2006 23:45:00 +0000</pubDate>
    </item>

    <item>
      <title>Few online players for some popular games</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060226b</link>
      <description>
	There are really few online player for some popular
	games. I guess that they are all playing WoW.
      </description>
      <category>videogames</category>
      <category>online</category>
      <category>MMORPG</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060226b</guid>
      <pubDate>Sun, 26 Feb 2006 23:30:00 +0000</pubDate>
    </item>

    <item>
      <title>KIAX keeps improving</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060226</link>
      <description>
	KIAX version 0.8.5 is more polished and useful. Too bad
	ALSA support is still missing.
      </description>
      <category>VoIP</category>
      <category>IAX2</category>
      <category>KIAX</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060226</guid>
      <pubDate>Sun, 26 Feb 2006 23:20:00 +0000</pubDate>
    </item>

    <item>
      <title>Firefox caches a bit too much</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060219b</link>
      <description>
	Firefox gets mocked a bit for memory leaking, in
	particular an intentional one, which however is misguided...
      </description>
      <category>web</category>
      <category>browser</category>
      <category>Mozilla</category>
      <category>Firefox</category>
      <category>Konqueror</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060219b</guid>
      <pubDate>Sun, 19 Feb 2006 21:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Reiser4 seminar at Google Talks</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060219</link>
      <description>
	Found an interesting video of a seminar about the design
	the the Reiser4 file system, and the aim to escape
	purely hierarchical or tabular arrangements.
      </description>
      <category>file system</category>
      <category>naming</category>
      <category>Linux</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060219</guid>
      <pubDate>Sun, 19 Feb 2006 17:10:00 +0000</pubDate>
    </item>

    <item>
      <title>PS/3, GPU and software anti-aliasing</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060212</link>
      <description>
	An informal interview with a PS/3 developer reveals that
	many PS/3 will not manage 1080p except via upscaling,
	and that some people have been doing software anti-aliasing
	on it, presumably using one or more SPEs.
      </description>
      <category>PS/3</category>
      <category>videogames</category>
      <category>parallelism</category>
      <category>CPU</category>
      <category>GPU</category>
      <category>graphics</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060212</guid>
      <pubDate>Sun, 12 Feb 2006 22:30:00 +0000</pubDate>
    </item>

    <item>
      <title>Updated KDE startup script</title>
      <link>http://WWW.sabi.co.UK/#sourcesDokde</link>
      <description>
	Updates and fixes to my version of the KDE startup script.
      </description>
      <category>Linux</category>
      <category>X11</category>
      <category>KDE</category>
      <category>configuration</category>

      <pubDate>Sun, 12 Feb 2006 21:55:00 +0100</pubDate>
    </item>

    <item>
      <title>Intel on ELF linking optimisation</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060204b</link>
      <description>
	An Intel paper discusses ELF performance improvements,
	and in particular the ''visibility'' issues, with some details.
      </description>
      <category>ELF</category>
      <category>Linux</category>
      <category>Intel</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060204b</guid>
      <pubDate>Sat, 04 Feb 2006 15:45:00 +0000</pubDate>
    </item>

    <item>
      <title>ECC RAM and recent motherboards with ECC</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060204</link>
      <description>
	Which motherboards do ECC given ECC RAM? Not many, and
	here are some pointers by chipset and by model.
      </description>
      <category>hardware</category>
      <category>RAM</category>
      <category>memory</category>
      <category>ECC</category>
      <category>motherboard</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060204</guid>
      <pubDate>Sat, 04 Feb 2006 15:25:00 +0000</pubDate>
    </item>

    <item>
      <title>Intel, Core Duo, and AMD price umbrella</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060203</link>
      <description>
	it is not clear why Apple switched to Intel, as the Core
	Duo is a pretty good chip for laptops. It is also
	interesting that Intel is playing low price alternative
	to AMD's leading dual core chips.
      </description>
      <category>hardware</category>
      <category>Apple</category>
      <category>Intel</category>
      <category>dual core</category>
      <category>CPU</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060203</guid>
      <pubDate>Fri, 03 Feb 2006 22:30:00 +0000</pubDate>
    </item>

    <item>
      <title>PS/3, ray tracing, Cell and NVIDIA</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060201</link>
      <description>
	Ray tracing is another possible use of those 7 SPEs in
	a PS/3. Remembering the Transputer. And what about
	in-game movies?
      </description>
      <category>PS/3</category>
      <category>simulation</category>
      <category>videogames</category>
      <category>parallelism</category>
      <category>CPU</category>
      <category>GPU</category>
      <category>graphics</category>
      <category>ray tracing</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060201</guid>
      <pubDate>Wed, 01 Feb 2006 22:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Reduced spin-up current options</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060130</link>
      <description>
	Just discovered that a manufacturer offers a reduced
	spin-up current option for 2.5" drives meant to be used
	in external USB/FW2 boxes.
      </description>
      <category>hard disc</category>
      <category>PSU</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060130</guid>
      <pubDate>Mon, 30 Jan 2006 22:20:00 +0000</pubDate>
    </item>

    <item>
      <title>Lack of security is like lack of ECC for RAM</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060129c</link>
      <description>
	As a smart friend points out, lack of security measures
	gives the same false sense of security as lack of ECC
	for RAM.
      </description>
      <category>security</category>
      <category>auditing</category>
      <category>memory</category>
      <category>RAM</category>
      <category>ECC</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060129c</guid>
      <pubDate>Sun, 29 Jan 2006 17:30:00 +0000</pubDate>
    </item>

    <item>
      <title>Another reason why RAID5 is not a good idea</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060129b</link>
      <description>
	As a smart friend points out, RAID5 also implies a lot
	more disk operations for the same amount of data, which
	can mean a lot more wear and tear of the hard discs.
      </description>

      <category>storage</category>
      <category>system administration</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060129b</guid>
      <pubDate>Sun, 29 Jan 2006 17:25:00 +0000</pubDate>
    </item>

    <item>
      <title>WINE often faster than MS Windows</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060129</link>
      <description>
	Some benchmarks shows that WINE under GNU/Linux is often
	as good or better than MS Windows at running WIN32 programs.
      </description>
      <category>Linux</category>
      <category>GNU</category>
      <category>WIN32</category>
      <category>WINE</category>
      <category>WINE</category>
      <category>OpenGL</category>
      <category>benchmark</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060129</guid>
      <pubDate>Sun, 29 Jan 2006 17:15:00 +0000</pubDate>
    </item>

    <item>
      <title>Growth of MMORPG sales and decline of PC game sales</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060128c</link>
      <description>
	Looking at an impressive chart of the number of MMORPG
	players over time suggests that the MMORPG industry is
	taking a lot of customers and revenue from the PC game
	industry.
      </description>
      <category>games</category>
      <category>MMORPG</category>
      <category>PC</category>
      <category>sales</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060128c</guid>
      <pubDate>Sat, 28 Jan 2006 18:35:00 +0000</pubDate>
    </item>

    <item>
      <title>Gran Turismo. PS/3 and partitioning by actor</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060128b</link>
      <description>
	Sony game consoles are probably designed for the benefit
	of their signature games, primarily Grant Turismo, and
	this may mean that the PS/3 is designed for partitioning
	load by actor, not by global effect or game pipeline step.
      </description>
      <category>games</category>
      <category>PS/3</category>
      <category>parallelism</category>
      <category>SMP</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060128b</guid>
      <pubDate>Sat, 28 Jan 2006 17:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Games push PC technology, and why</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060128</link>
      <description>
	The driver of most upgrades to my PC has been increased
	game requirements. 
      </description>
      <category>games</category>
      <category>GPU</category>
      <category>SMP</category>
      <category>technology</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060128</guid>
      <pubDate>Sat, 28 Jan 2006 17:20:00 +0000</pubDate>
    </item>

    <item>
      <title>RAID10 vs. RAID5, one data point</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060127b</link>
      <description>RAID10 is over 10 times faster than RAID5 for backups,
	an even more dramatic difference than expected.</description>

      <category>RAID5</category>
      <category>RAID10</category>
      <category>disk</category>
      <category>backup</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060127b</guid>
      <pubDate>Fri, 27 Jan 2006 21:30:00 +0000</pubDate>
    </item>

    <item>
      <title>Now I have 1GiB RAM, and yes it is much better</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060127</link>
      <description>
	I have sold out, or rather bought in 1GiB RAM, and
	things are so much better than with 512MiB that it is
	likely most kernel/KDE/GNOME developers have at least
	1GiB themselves.
      </description>
      <category>Linux</category>
      <category>RAM</category>
      <category>ECC</category>
      <category>1GiB</category>
      <category>swapping</category>
      <category>virtual memory</category>
      <category>KDE</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060127</guid>
      <pubDate>Fri, 27 Jan 2006 21:05:00 +0000</pubDate>
    </item>

    <item>
      <title>Photo of the internals of a high speed 3.5" disc drive</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060113</link>
      <description>
	Found a nice photo of a high speed 3.5" disc drive which
	shows it actually has 2.5" platters.
      </description>
      <category>hard disk</category>
      <category>hard drive</category>
      <category>performance</category>
      <category>ATA</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060113</guid>
      <pubDate>Sat, 13 Jan 2006 12:15:00 +0000</pubDate>
    </item>

    <item>
      <title>Found extensive test of compression programs</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060111</link>
      <description>
	A friend has sent me a link to an extensive test of
	compression programs, which includes decompression speed.
      </description>
      <category>Linux</category>
      <category>GNU</category>
      <category>compression</category>
      <category>decompression</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060111</guid>
      <pubDate>Wed, 11 Jan 2006 22:10:00 +0000</pubDate>
    </item>

    <item>
      <title>Trends in game development</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060105c</link>
      <description>
	All games are becoming mods, and some games are
	developed almost only to showcase a game engine which is
	the real source of profits for the game studio
	developing both game and engine.
      </description>
      <category>games</category>
      <category>marketing</category>
      <category>game engine</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060105c</guid>
      <pubDate>Fri, 05 Jan 2006 21:45:00 +0000</pubDate>
    </item>

    <item>
      <title>Another interview with John Carmack</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060105b</link>
      <description>
	This interview is rather expansive and contains a
	discussion of PC game piracy (with some dodgy arguments)
	and on console development: if it looks like a PC it is good.
      </description>
      <category>games</category>
      <category>mobile phones</category>
      <category>Xbox 360</category>
      <category>PS/3</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060105b</guid>
      <pubDate>Fri, 05 Jan  2006 20:40:00 +0000</pubDate>
    </item>

    <item>
      <title>Interview with John Carmack</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060105</link>
      <description>
	John Carmack comments on mobile phone games and toolkits
	for them.
      </description>
      <category>games</category>
      <category>mobile phones</category>
      <category>Doom</category>
      <category>BREW</category>
      <category>J2ME</category>
      <category>Java</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060105</guid>
      <pubDate>Fri, 05 Jan  2006 20:15:00 +0000</pubDate>
    </item>

    <item>
      <title>Useful text mode/command line programs</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060102</link>
      <description>
	Found a page with a list of preferred command line
	programs, so some comments and additions.
      </description>
      <category>curses</category>
      <category>web</category>
      <category>shell</category>
      <category>command line</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060102a</guid>
      <pubDate>Mon, 02 Jan 2006 16:05:00 +0000</pubDate>
    </item>

    <item>
      <title>The Linux registry, and other horrors</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060101</link>
      <description>
	When buyer fail to appreciate the importance of a
	feature of a product, like maximum spin up current or
	longest seek time, the feature stops being mentioned in
	the specifications and then itself becomes worse. For
	example this has happened to ECC support in motherboard
	chipsets, which is a very important and nearly free feature.
      </description>
      <category>hard drive</category>
      <category>performance</category>
      <category>ECC</category>
      <category>RAM</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060101</guid>
      <pubDate>Sun, 01 Jan 2006 16:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Specifications become invisible and then the
	feature fades too</title>
      <link>http://WWW.sabi.co.UK/blog/anno06-1st.html#060101</link>
      <description>
	When buyer fail to appreciate the importance of a
	feature of a product, like maximum spin up current or
	longest seek time, the feature stops being mentioned in
	the specifications and then itself becomes worse. For
	example this has happened to ECC support in motherboard
	chipsets, which is a very important and nearly free feature.
      </description>
      <category>hard drive</category>
      <category>performance</category>
      <category>ECC</category>
      <category>RAM</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno06-1st.html#060101a</guid>
      <pubDate>Sun, 01 Jan 2006 15:15:00 +0000</pubDate>
    </item>

    <item>
      <title>More on which hard drives to get for an upgrade</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051227b</link>
      <description>
	More thinking on the various conflicting requirements
	for my upgrade main drive and its 3 backup drives. The
	GHST and Maxtor drives match my requirements for both
	internal and external drives, but I also want
	diversification of manufacturer.
      </description>
      <category>hard drive</category>
      <category>performance</category>
      <category>storage</category>
      <category>backup</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051227b</guid>
      <pubDate>Tue, 27 Dec 2005 18:20:00 +0000</pubDate>
    </item>

    <item>
      <title>Poor swap speed with Linux</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051227</link>
      <description>
	It happens fairly often that my Linux desktop has to
	swap in processes paged out because of memory pressure,
	and this happens at 1MiB/s on a disc capable of 40MiB/s,
	because of lots of arm movement. No relief in sight.
      </description>
      <category>Linux</category>
      <category>performance</category>
      <category>virtual memory</category>
      <category>swapping</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051227</guid>
      <pubDate>Tue, 27 Dec 2005 18:00:00 +0000</pubDate>
    </item>

    <item>
      <title>Using only the outer cylinders of a cheap disc</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051226e</link>
      <description>
	To enjoy fast disc performance one can buy a large
	desktop disc and then use only its outer cylinders.
      </description>
      <category>hard drive</category>
      <category>performance</category>
      <category>server</category>
      <category>DBMS</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051226e</guid>
      <pubDate>Mon, 26 Dec 2005 23:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Recent PC games require top of the line systems</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051226d</link>
      <description>
	Having played some not so recently released FPS games on
	my average PC they end up with low frame rates.
      </description>
      <category>graphics</category>
      <category>GPU</category>
      <category>game</category>
      <category>FPS</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051226d</guid>
      <pubDate>Mon, 26 Dec 2005 23:25:00 +0000</pubDate>
    </item>

    <item>
      <title>Power consumption of current hard drives</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051226c</link>
      <description>
	Hoping to upgrade my desktop PC to some larger discs, I
	consider also spin-up power draw as some of those discs
	will end up as backup discs in an external USB case. Only
	a few models can do.
      </description>
      <category>hard drive</category>
      <category>USB2</category>
      <category>Firewire</category>
      <category>PSU</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051226c</guid>
      <pubDate>Mon, 26 Dec 2005 22:50:00 +0000</pubDate>
    </item>

    <item>
      <title>Latest numbers in over time filesystem slowdown</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051226b</link>
      <description>
	Four weeks ago I reinstalled from scratch my Linux root
	partition, and I have been regularly upgrading it. How
	much slower overall has it become? Only 1.8 times slower
	with JFS.
      </description>
      <category>filesystem</category>
      <category>Fedora</category>
      <category>speed</category>
      <category>JFS</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051226b</guid>
      <pubDate>Mon, 26 Dec 2005 22:10:00 +0000</pubDate>
    </item>

    <item>
      <title>Quake 4 performance on dual core systems</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051226</link>
      <description>
	Some comments on an interesting article as to
	performance of a Quake 4 update that is multithreaded to
	take advantage of hyperthreading or multiple CPUs, and
	some comments on general strategies to achieve good speedups.
      </description>
      <category>game</category>
      <category>Quake 4</category>
      <category>hyperthreading</category>
      <category>SMP</category>
      <category>Intel</category>
      <category>AMD</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051226</guid>
      <pubDate>Mon, 26 Dec 2005 22:05:00 +0000</pubDate>
    </item>

    <item>
      <title>Using 'ext2' under MS Windows works well</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051219</link>
      <description>
	Having switched from FAT32 to 'ext2' the filesystems I
	share between MS Windows and GNU Linux on my home PC a
	few weeks ago, so far I am quite happy with it.
      </description>
      <category>Linux</category>
      <category>MS Windows</category>
      <category>FAT32</category>
      <category>ext2</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051219</guid>
      <pubDate>Mon, 19 Dec 2005 22:05:00 +0000</pubDate>
    </item>

    <item>
      <title>To optimize do not use expression oriented APIs</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051217</link>
      <description>
	When optimizing vector codes especially on CPUs with SIMD
	oriented operations it is quite useful to avoid
	expression oriented constructs like 'x = y + z', and just
	use ''and becomes'' oriented APIs, like 'x = y; x += z'.
      </description>
      <category>vector</category>
      <category>architecture</category>
      <category>optimization</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051217</guid>
      <pubDate>Sat, 17 Dec 2005 17:15:00 +0000</pubDate>
    </item>

    <item>
      <title>memory locality studies are not dead</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051216</link>
      <description>
	I have found a book that contains a number of mostly
	well written essays on improving memory reference
	locality. It is amazing that however little this is
	practiced it is not yet a totally forgotten area.
      </description>
      <category>optimization</category>
      <category>architecture</category>
      <category>locality</category>
      <category>memory</category>
      <category>cache</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051216</guid>
      <pubDate>Fri, 16 Dec 2005 22:10:00 +0000</pubDate>
    </item>

    <item>
      <title>More stunning filesystem performance discoveries!</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051204</link>
      <description>
        Very interesting and unexpected performance issues with
        filesystems, in particular 'ext3' with the 'dir_index'
        option.
      </description>
      <category>UNIX</category>
      <category>Linux</category>
      <category>Fedora</category>
      <category>filesystem</category>
      <category>ext3</category>
      <category>JFS</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051204</guid>
      <pubDate>Sun, 04 Dec 2005 21:45:00 +0000</pubDate>
    </item>

    <item>
      <title>Some decompression speed tests</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051203</link>
      <description>
	Comparison of 'bunzip2', 'gunzip' and 'lzop -d' as to
	decompressing archives containing the same data.
      </description>
      <category>UNIX</category>
      <category>Linux</category>
      <category>lzop</category>
      <category>bunzip2</category>
      <category>gunzip</category>
      <category>compression</category>
      <category>decompression</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051203</guid>
      <pubDate>Sat, 03 Dec 2005 20:10:00 +0000</pubDate>
    </item>

    <item>
      <title>KDE 3.5 has been released, Fedora 4 RPMs available</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051202b</link>
      <description>
	After the release of KDE 3.5, some Fedora 4 RPMs are now
	available.
      </description>
      <category>Linux</category>
      <category>RedHat</category>
      <category>Fedora</category>
      <category>RPM</category>
      <category>KDE</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051202b</guid>
      <pubDate>Fri, 02 Dec 2005 18:55:00 +0000</pubDate>
    </item>

    <item>
      <title>Doom III with Voodoo 2 SLI, and bump mapping</title>
      <link>http://WWW.sabi.co.UK/blog/anno05-4th.html#051202</link>
      <description>
	Some people have run Doom III on a Voodoo 2 SLI graphics
	card, which does not have bump mapping. This makes clear
	just how simple is the geometry in Doom III and how
	effective bimp mapping at adding detail.
      </description>
      <category>games</category>
      <category>Doom III</category>
      <category>graphics</category>
      <category>OpenGL</category>

      <guid isPermaLink="true"
        >http://www.sabi.co.uk/Notes/anno05-4th.html#051202</guid>
      <pubDate>Fri, 02 Dec 2005 17:50:00 +0000</pubDate>
    </item>
  </channel>

  <script xmlns="http://www.w3.org/1999/xhtml" type="text/javascript"
    src="style/rss.js"></script>
<!--
  ;;; Local Variables:			***
  ;;;   tab-width:		8	***
  ;;;   indent-tabs-mode:	t	***
  ;;;   fill-column:		72	***
  ;;; End:				***
  vim:set noet ts=8 tw=72 sw=2:
-->
</rss>
