[Axiom-developer] RE: comment

[Bill Page]

On April 16, 2007 1:09 PM Tim Daly wrote:
> 
> As I recall the whole Axiom project spent a considerable amount
> of effort trying to resolve the SVN problem. You did a great
> job trying to set up SVN on axiom-developer (successfully?) and
> Google (unsuccessfully). 

I never completed the full setup of the SVN server component on
axiom-developer because that required an upgrade to Apache and
our current Apache configuration file does not work automatically
under the new version of Apache. Some small changes are required
and more testing, but I did not complete this yet. It no longer
seemed important once the problems with SourceForge were solved.

The problem with Google is different because we kept tripping
up on the repository size constraint even after they increased
the limit by about a factor of 4. This could be solved if we
decided to abandon a lot of the history which involves a lot
of unnecessary duplication etc. Again this could be done if any
really had any interest in it.

> 
> I eventually experimented with other systems. You advocated
> darcs and mercurial as I remember. I looked at darcs but my
> investigations, and I believe your comments, left me with the
> impression that it is not ready for a full, large projects.
> I did not try mercurial.

In fact both are used on projects about the size of Axiom and we
have both darcs and mercurial mirrors of the build-improvements
branch on axiom-developer.org.

> 
> SVN times out on me and I have to do a checkout+12update cycle
> to get a full axiom copy. Checkin takes over half an hour. SVN
> uses twice the space of the Axiom distribution. It all feels
> slow, ponderous, and heavy. And SVN wedges, demanding I "cleanup",
> which never succeeds.
>

That was exactly my own experience with SVN up until December
of last year. At that time SourceForge provided a new url for
SVN access that solved the problem. See:

http://lists.nongnu.org/archive/html/axiom-developer/2006-12/msg00156.html

I have no problems using SVN on Linux and on Windows for either
of the locations where I work on Axiom since that time.
 
> I eventually tried git. Checkout and checkin over the network
> take about the same time as a local disk-to-disk copy. It uses
> only a tiny amount of additional disk space. I like it so much
> that I've moved all of my research work into git and use it for
> normal work purposes.

>From my experience, git and mercurial are very similar in speed,
storage requirements and even commands. darcs is slower (not that
much slower) and provides much finer grain patch control for
merge and for cherry-picking.

> 
> So, yes, you did reproduce my SVN problems and you discussed how
> to solve my problems. However, SVN itself has problems (slow,
> unreliable, and large) which have no solution. The question
> wasn't "can it be made to work" but rather "why waste the time
> and space".
>

I do not think it is slow or unreliable since we got the new url
from SourceForge, but of course we cannot do anything about it's
size. The main reason to put up with the waste of time and space
is for the sake of cooperation with other developers who just as
unwilling to change. Somebody has to compromise or the result is
the current situation.
 
> Git is the best technical solution, thus I use git. 
> To me, that's progress.
> 

To me, progress occurs when we manage to cooperate and work
smoothly together to achieve a common goal.

Regards,
Bill Page.