Whygitisbetterthan.com, some advocacy from GitHub

[Martin Pool]

On Tue, Dec 2, 2008 at 7:28 PM, Eugene Wee <crystalrecursion at gmail.com> wrote:
> Hi,
>
> I think you missed out "Git is small" as the 4th area. I am not sure what to
> make of it though. Another comparison made just over half a year ago had
> Bazaar win in this department, so did Git make a huge improvement in this
> area over the past half a year, or was the comparison (unintentionally)
> rigged in Git's favour?

In the respective current formats, git re-deltas files when repacking
and we do not, we just reorganize and reindex the existing deltas
(iiuc).  We're working on rediffing them in the brisbane-core project,
and that might be where the numbers came from.

> Also, did the author detect a bug? I note the comment that "the second Bzr
> number is after I ran 'bzr pack', which I thought would make it smaller, but
> ended up making it much, much larger for some reason."

The pack command does not delete the just-obsoleted pack files, in
case they're still in use by eg an http client.   In fact, I think
they're never automatically deleted at present, and they should be
<https://lists.ubuntu.com/archives/bazaar/2008q1/036740.html>
<https://bugs.edge.launchpad.net/bzr/+bug/304320>

The one I find most unlikely (and may be inadvertent measurement
error) is actually the working tree operations like status and diff,
where he claims over 14s and for me it's under 300ms.  That's in a new
copy of that django branch, using bzr 1.10, with a small change to one
file.  Tag is about 200ms.  I would guess maybe he's running from
python source without the extensions built.

-- 
Martin <http://launchpad.net/~mbp/>