From: Dominik Riebeling <dominik.riebeling_at_gmail.com>
Date: Wed, 15 Sep 2010 19:01:19 +0200

On Wed, Sep 15, 2010 at 6:38 PM, Paul Louden <paulthenerd_at_gmail.com> wrote:
> I guess my question is, do we want to ask people to *try* for useful commit
> messages, or do we want to say "if we can edit them later, anything is
> okay?"

I'm not sure if the latter would happen since changing the commit
message also creates noise (and you basically publish a "I did it
wrong" message, and nobody likes to be wrong, right?).

I'm using git-svn a lot and that allows me to review every change
before actually committing it to the main repository. Basically, my
current workflow usually looks like this:
- create a feature branch
- work, do several commits
- when done, rebase the feature branch, then squash it to a single commit
- review the change, sometimes even on a different branch. Since it's
a single commit I can easily amend the commit message of that commit
until I'm satisfied. It did happen that I amended the commit message
several times until I was satisfied. It even did happen that I noticed
issues with the commit during that phase (like forgotten files for
which using a different branch is quite helpful), so I can fix that.
- dcommit it.

I don't know of any numbers but I'm under the impression that quite a
lot of committers use git-svn (or mirror with another DVCS). And if
you use such a tool I don't see much excuses for making lazy commits.
I for myself tend to rather make a commit the next day than to make a
broken one (though I'm not always sucessful at that).


 - Dominik
Received on 2010-09-15