<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0" xmlns:admin="http://webns.net/mvcb/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:creativeCommons="http://backend.userland.com/creativeCommonsRssModule" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:html="http://www.w3.org/1999/html" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:slash="http://purl.org/rss/1.0/modules/slash/"><channel><title>Ross Burton</title><link>http://www.burtonini.com/blog</link><description>A potted account of Ross' life</description><language>en</language><ttl>60</ttl><dc:creator>Ross Burton</dc:creator><admin:generatorAgent rdf:resource="http://pyblosxom.sourceforge.net/"/><admin:errorReportsTo rdf:resource="mailto:ross@burtonini.com"/><item><title>EDS DBus vs CORBA</title><guid isPermaLink="false">computers/eds-dbus-2005-12-15-14-54</guid><link>http://www.burtonini.com/blog/computers/eds-dbus-2005-12-15-14-54</link><description>Ah, Novell have released an initial report comparing the DBus and CORBA ports of EDS . As the author of ...</description><content:encoded><![CDATA[    <p>
      Ah, Novell have released an initial report comparing the <a
      href="http://go-evolution.org/DBus_Port_of_EDS">DBus and CORBA ports of
      EDS</a>.  As the author of the DBus port, I should make some observations.
    </p>
    <p>
      The incorrect reporting of large numbers of contacts in a book view is
      very odd, and I've not seen that before.  I'll try and replicate that
      locally, I did change some message limits so maybe they are being hit.
    </p>
    <p>
      Occassional <tt>dbus_g_proxy_cancel_call</tt> assertions are known to
      happen. I've a potential fix locally, but it requires a patch to DBus,
      which hasn't been committed yet.
    </p>
    <p>
      Now to the benchmarks.  The primary optimisation target for my work has
      been in the book views, as these are the primary bottleneck in a
      real-world application.  This is visible in Test 1, where the DBus port
      takes 17 seconds compared to 72 seconds for CORBA.  A number of the
      optimisations I've made can be applied directly to the CORBA code too, and
      are currently sitting in Bugzilla.
    </p>
    <p>
      Test 2, calling <tt>e_book_get_contacts()</tt> is interesting.  It may be
      that verbose mode is enabled in DBus, which has a serious performance
      impact on large messages (and this is a very large message).
      Alternatively this difference may simply be due the message bus, which
      adds a layer of IPC (and thus string copies) which CORBA doesn't have.  As
      very few real-world applications (I'd say none) use this function (it
      blocks the interface unlike book views which are asynchronous), this isn't
      a very meaningful benchmark in the scheme of things.
    </p>
    <p>
      For the calendar, there has been no performance turning done at all.  It's
      known that the calendar port to DBus isn't quite complete, and needs
      cleaning up and performance work.
    </p>
]]></content:encoded><dc:date>2005-12-15T14:54:00Z</dc:date></item></channel></rss>