Re: [tlug] Open-source repository question

["Stephen J. Turnbull" <stephen@example.com>]

Edward Middleton writes:
 > Stephen J. Turnbull wrote:
 > > Edward Middleton writes:
 > >  > If you are hoping to pull together changes from some of the
 > >  > variants or make it easier for others to do so then git is probably
 > >  > the best tool
 > >
 > > Been there, done that, and no, git is not clearly the best tool for
 > > that.

 > [That] wasn't my point.  Git is better then darcs because more
 > people know git then darcs.

But it was precisely my point that that logic depends on assumptions
about Jim's goal.  If Jim's goal is to collect a bunch of active
hackers and revitalize the project, then using git will lower the
barrier to such developers, as you point out.  But if his goal is
simply to merge the versions and leave the product in a public place,
then he may prefer a VCS with a UI oriented to selectively committing
pieces of a merge as a coherent changeset.  Even if that VCS is not
very popular.

Also, I doubt Jim is an OSS OS zealot.  Windows users dislike git, so
git is not necessarily the best choice even if communication among
developers is the goal.

 > If you are interested in selling a better DVCS then use darcs, if
 > you are interested in distributed development use Git because if
 > you use darcs you will spend your time selling darcs.

Well, not necessarily.  You can also choose to spend your time using
Darcs, and not selling anything.  I don't recommend Darcs as a VCS for
active, multideveloper projects, but it has certain nice features for
one-man, not-so-active projects.

 > >  > My general feeling about sourceforge et. al. is that they are places
 > >  > where projects go to die.
 > >
 > > By comparison to github, that's not surprising IMO.  (1) SourceForge
 > > has the longest history, but never deletes a project AFAIK.  Of course
 > > it accumulates dead ones.  (2) github is going to attract developers
 > > whose projects are still at the frenetic stage of development.  Of
 > > course it's going to look lively.
 > 
 > I think it has more to do with good tools and supporting a more
 > effective development model.

Maybe; IMO the jury's still out on that.  The conclusion of PEP 374
(http://docs.python.org/dev/peps/pep-0374/) was that in terms of power
and development models, there's little to chose from among git, hg,
and bzr.  In the end, bzr got axed because performance sucked on all
platforms, and hg edged out git on the basis of git's poor support for
Windows and confusing UI.  (Contrary to popular belief, unlike
Haskell's use of Darcs, dogfooding was never an important issue for
Python.)