Bazaar performance

Teemu Likonen tlikonen at iki.fi
Sat May 17 07:58:40 BST 2008


John Arbash Meinel wrote (2008-05-16 13:40 -0500):

> The other thing I wonder about is whether running "bzr pack" to
> compact down the repository would help. Right now we have some scaling
> issues as the number of pack files increases. We are looking into it,
> but often a "bzr pack" makes things noticeably faster.

Hmm, I wonder what this means in BzrVsGit document:

    To maintain performance, Git repositories require packing. The pack
    technology recently added to Bazaar is smart enough to do this for
    you. You can pack when you want but if you don't, things will still
    run well over time.[1]

I may be misinterpreting something but I have to admit that the above
text seems like a joke in the light of recent performance tests (see [2]
for example).

Git is also "smart enough" to do the packing and garbage collecting
automatically. Some of the "smartness" can also be fine-tuned by user.
The recent discussion[3] under the subject "Bazaar experimental branch
removal" suggests that bzr lacks garbage collecting functionality in
shared repositories. I know it can be worked around by creating new
shared repo etc. (and it's enough for me), but git still seems much
smarter in the garbage collecting too with all the reflogs and object
expire times.

So to me it really seems that the above quoted text is not at all an
argument for bzr. It even looks like an irony. I have no problems with
open advocacy. I just want to point out that the above text may actually
turn against bzr if people see it as a bad joke or a lie.


-----------------
1. http://bazaar-vcs.org/BzrVsGit#head-664a45b37d0306192b11fcfb5cfb529b17bc712e
2. http://thread.gmane.org/gmane.comp.version-control.bazaar-ng.general/38404/focus=38960
3. http://thread.gmane.org/gmane.comp.version-control.bazaar-ng.general/41654/focus=41687



More information about the bazaar mailing list