Re: Switch to git after all?
Joey Hess <email@example.com> writes:
> I advise against using git in some nonstandard way, such as only
> checking debian/ into the debian git repositories. There is a very
> wide range of tools for managing packages with git, and making this
> choice will close off using many of those tools, or require using them
> in nonstandard ways, and so eliminate many of the possible positive
> effects of switching to git in the first place.
tl;dr: What he said.
I quit being an actual DD a decade ago, so it would be of no surprise to
me if my views were outdated or irrational or just plain wrong, and thus
I'm a little reluctant to speak up. Especially since I deeply
appreciate the work everyone here has done to maintain the Haskell
ecosystem in Debian, and would not wish to appear to be denigrating
their effort and achievement.
Still, I have found the way the haskell packages are currently
maintained very unintuitive, and that has served as a significant enough
barrier that although I maintain packages for a dozen or so libs I need
for a pesonal project, the couple of times I've considered doing some
work around the periphery to help, I have inevitably dropped the idea.
Some of my problem was surely darcs---not *just* that it wasn't git
(though that was part of it), but that when I went to the trouble to
figure out what I needed to do, I usually had to, say, run the same
update command multiple times before it would succeed.
So I wholeheartedly support moving to git; that would remove one
Beyond that, though, I find the only-maintain-debian/-in-your-vc model
strange---and I began as a DD 18 years ago, when we had basically none
of the current package infrastructure, so it's not like I don't know how
to do it without fancy tools; I remember when 'fakeroot' was the big
innovation because you didn't actually have to be root---and it will
still represent some barrier for my participation, at least.
On the other hand, yesterday I updated my package for the thyme library,
and it took me roughly 3 minutes of actual work. Between debian/watch
and a couple of dumb git aliases I find I can keep up with very little
effort. My experience maintaining 50+ Perl libraries, and something
like 170+ Drupal modules is similar.
What it comes down to is that if the haskell libs were being maintained
in git, using the de-facto tools for doing so, I would be looking to
re-activate my Developer status and make an active contribution.
To be clear: that notional commitment shouldn't outweigh Joachim's
comfort level, given his overwhelming contribution...but I do wonder
whether the need for that overwhelming level of contribution wouldn't be
lessened if the packages were maintained more like others in the Debian
ecosystem. Perhaps that's a 'Worse is better' argument, though.