In an earlier comment on the tread, someone said he didn't care about the relative performance of git vs bzr.
If you want to encourage a larger development community, performance should be considered. Becasue of python bzr is considerably more VM hungry than git; in a large repo like inkscape (over 22222 svn revs) that will be an issue for many potential contributors.
As an example of relative memory pressure, a git-svn conversion of the full inkscape repo takes about 6 CPU-core hours on a 3GHz core2 with ample RAM, and needs about two Gigs of that ram (including both the process' VM and the filesystem cache) to avoid disk I/O wait. Using bzr-svn to do that needs around two Gigs just for the process, plus another two Gigs or more for the cache. (The bzr-svn conversion isn't finished yet, the resource usage could end up even larger.)
Also, note that if git is chosen, those who prefer bzr can use the git plugin for bzr to interact with the git repo.
Finally, orthogonal to the question of which DVCS, given this:
,----< svn ls https://inkscape.svn.sourceforge.net/svnroot/inkscape > | Booleans_article/ | CVSROOT/ | Makefile_insert/ | branding/ | clipartbrowser/ | css_inline/ | dms/ | doc-docbook/ | experimental/ | gsoc-testsuite/ | inkscape/ | inkscape_drupal/ | inkscape_marketing/ | inkscape_project/ | inkscape_testfiles/ | inkscape_web/ | ps3convert/ | svg_metadata/ | svgtopdf/ | user_manual/ `----
Each of those 19 directories (excluding CVSROOT) should be converted into its own git or bzr repo. This will both keep the repos more managable (I'll test out the per-subdir conversions) and avoid needless resource usage by those who use them.
For git, sf enables having several repos in a project; I'm sure lp does just as well as that for bzr.
-JimC