axiom-developer
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Axiom-developer] GIT


From: root
Subject: Re: [Axiom-developer] GIT
Date: Wed, 25 Oct 2006 10:21:22 -0400

> > i've been unable to use SVN and could generate no interest in GIT.
> 
> It's interesting to me, but I'm somewhat dismayed by the multitude of
> systems we are going through.  I would have been happy to standardize
> on and learn Arch, but as it is no longer maintained that's not really
> a viable long term option.
> 
> If GIT works fast and suits our needs, I'd be glad to learn that.  More
> important to me is that we pick SOMETHING.  Automatically syncing to
> other repositories is fine for backup purposes, but I think it is a
> distraction to be spending so much time on tools rather than Axiom
> itself.
> 
> There are too many "live" trees.  I can understand different branches
> on an individual machine, but for online work of this nature I think
> scattering changes over half a dozen repositories is a Bad Idea.  Too
> much chance of a patch getting lost.  I uploaded some files that had
> been broken down into smaller chunks onto the wiki axiom--test--1
> branch (I think).  Did that ever get noticed or did it just get lost? 
> It's too easy to do work that doesn't get noticed by the "cricical core
> mass" of the project.  Maxima has one central repository where all
> changes wind up, and I think that has worked out fairly well.  Branches
> are generally made for releases only.
> 
> I'm less concerned with what tool we use, and more concerned that we
> just pick a tool and use it.  Ultimately everyone can generate tarballs
> and use tools to make and merge changesets into their own favorite
> local system - for the project, we need to focus more and avoid the
> endless tool parade.  I would have been fine with SVN, but the problems
> key developers are having with running it make it problematic.  If
> everybody likes GIT that's great - I'll be glad to install it - but
> let's pick something and go with it.


the procedure for changing axiom is well known.

do the work (as literate, well documented code), 
send me and the list a "diff -Naur goldfile newfile" for each change. 
follow up to make sure somebody saw it because email doesn't always work.




the gold branch is maintained as axiom--main--1 in arch.
the gold branch is mirrored at savannah in CVS.
the gold branch is mirrored at sourceforge in CVS.
the gold branch is released about every 2 months.

the silver branch is (now) maintained as axiom--silver--1 in arch.
the silver branch is updated as changes are accepted.

there are development branches maintained on sourceforge in SVN.
there are development branches maintained in arch.




Arch continues because it works.

CVS continues because it's the worldwide default tool.

SVN won't work for me so i have stopped trying to use it.
it should "just work" and it doesn't. Gaby likes it and it works
well for him so he uses it. Bill wrestled heroically with it.

darcs is easy but doesn't seem to scale and gets nonlinearly slower.

git works like magic, is blindingly fast, and uses very little
additional space. compared to git, the arch, cvs, darcs and svn
tools are slow and use much more space (207M vs 304M for axiom). 
however only one person responded to the experiment so that idea died.





"just pick a tool and use it" presumes controlling authority which
no one has, it presumes vision of the "right path" which is even 
less available, and it presumes that change and chaos are bad but
which is really the source of strength.

some days it's hard to keep up but the real value of open source
is the ability for everyone to "lead in your own direction". 
everyone has an opinion but working code always wins in the end.

chaos and complexity are necessary evils.

t






reply via email to

[Prev in Thread] Current Thread [Next in Thread]