Re: Maintenance question

[bulia byak <buliabyak gmail com>]

On Wed, 25 Aug 2004 15:46:48 +0200, Leonard den Ottolander
<leonard den ottolander nl> wrote:
> I already figured that trying to regulate things too much is probably
> counter productive at this stage of revitalizing the project. 

True!

> would prefer to see a little reluctance towards committing
> (incomplete/untested) patches too quickly. Give yourself a week to check
> your own work before committing. This is why I propose to put up patches
> for review, also by people with commit access. It's still possible to
> commit them after a while if no-one responded.

Yes, that sounds like an ideal policy: "silence is a sign of
approval". If you have concerns, voice them, otherwise it gets
committed by default.

> Waiting with committing until a piece of work is complete also makes the
> Changelogs more useful to work with. Instead of having to read through
> dozens of Changelog entries to get an impression of what changed it's
> easier to have a single (large) Changelog entry describing a whole
> patch. This is especially important if people need to touch the same
> part of the tree.

This is true, but it is also true that smaller commits make it easier
to find the exact point where something broke, by testing dated
checkouts. The point of CVS is to make changes easily traceable; by
lumping together many patches into one commit we lose this advantage.
So we need some sort of a balance here. In Inkscape, our policy is
simple: the CVS must compile and basically work at all times; other
than that it's up to the developers how big or small or big are their
commits. We also don't have the requirement that each commit must be
commented in the changelog; you can do several small commits over
several days, and then describe them in one descriptive changelog
entry. And if anyone needs to see comments for each small commit,
that's what CVS history is for.

> Personally I do not have a big problem with working with patches,
> assuming they will be reviewed and hopefully committed in a reasonable
> time. Frustration only creeps in if they are not reviewed within a
> reasonable time, or not at all. Roland and Pavel Shirshov have been very
> responsive in this respect, although I wonder if they can keep up the
> pace if they remain with only the two of them ;) .

Yes, I think 2 or 3 developers with CVS access is certainly too low
for a project like this. The best ratio of the committing developers
to all active developers can be argued, but I think it's obviously too
low now.

By the way we could contact the fork projects (such as Arpi with his
AMC) to see if they would be interested to merge their stuff back.
With AMC in particular, there was some talk of merging in 2003 but it
apparently didn't play out for some reason.