Hi, On Wed, 2007-01-17 at 16:57 -0500, D Bera wrote: > Michal Pryc and Steven Xusheng Hou of Sun compared the four popular > desktop indexing systems - JIndex, Tracker, Strigi and Beagle. > http://mail.gnome.org/archives/tracker-list/2007-January/pdfLkb0uuBAEw.pdf > > Its a nice comparison, including stats and charts. You might find it > interesting if you arent subscribed to the tracker mailing list.
Sadly, they are using Beagle 0.2.7 in their tests, a 7 month old version. :( In releases since then, we added support for wildcard searches, indexing of archives and their contents, and of course tons of memory use fixes. The doc was originally written in November and updated in December; it would be nice to see updated Beagle stats. Beagle fares pretty well overall. Indexing speed is good, the index is relatively small, and it's still the most featureful of the indexers out there. I do have a little beef with some of their statistics: referencing numbers from "top" for memory usage -- VmSize is an effectively useless statistic because it includes all mappings, whether they're read-only or not. (This is why X is so big in top: it maps in all of video memory.) Dirty, private maps from pmap would have been a more useful statistic. Also, Beagle doesn't do so well in "daemon startup", but how they measure this is entirely undefined. By and large the writers did their homework -- with Beagle, anyway. The only real inaccuracy I found with regards to Beagle was in saying that the number of results in Beagle was capped at 100 without a way to override this. While it can't be overridden in beagle-search, it can with beagle-query or the various programmatic APIs. Newer versions of beagle-search show the total number of matched documents, but in general this isn't a weakness; Google won't show you beyond the 1000th hit, either, but you'd never know that because you never paged that far. Anyway, it's a great resource to have, and I appreciate the work by Michal Pryc and Xusheng Hou. Joe _______________________________________________ Dashboard-hackers mailing list [email protected] http://mail.gnome.org/mailman/listinfo/dashboard-hackers
