> Perhaps a more realistic score board would be to publish the projects
> with the highest number of commits per month. Not exactly an accurate
> reflection of project activity since the contest would tend to foster
> lots of smaller, otherwise unnecessary, commits. But, it might be an
> interesting benchmark on SVN itself. Projects could strive for 10,000
> or 100,000 commits per month.
>
> Reid
My pair programming partner and I have done 530 commits in a little bit
less than a month and a half. It would not too many of us to reach
10K commits/month.
I'm from the "commit early and often" school of development:
I checkin changes early, nearly always as soon as they pass tests,
sometimes checking in broken stuff on a branch (not to be released).
(Unfortunately, Subversion branching is more painful than CVS
branching, because Subversion branching requires more Makefile
changes.)
Also, I use SVN (and used CVS in the past) as a sort of Poor Man's
replicated filesystem. Building on local disks in /tmp is 5X
faster than builds on our NFS systems, so I work in /tmp
and check in early and often so that the data is preserved
in the repository in case the Linux box I am working on crashes.
This also allows me to continue working across network outages.
In theory, a good network filesystem with local caching would be
better; in practice, local filesystems with VC providing the backup
works best given the present sorry state of affairs.
And, in general, a replicated / disconnectable filesystem really should
have a VC system underlying it for conflict management.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Wed May 19 04:46:45 2004