Re: [Axiom-developer] patches

[Ralf Hemmecke]

Hi Bill,

> > I would very much like that there is only ONE version, namely 
> > axiom--test--1 editable through the wiki and that version is always
> >  in sync with a branch (call it silver/branches/axiom--test--1) on 
> > our subversion repository.

> I rather dislike these weird "arch" names.

No problem, but the branch on the SVN repository and the name you put
onto the website should be identical.

> Perhaps we can define a new branch in Silver specifically for this 
> purpose?

Yes, exactly.

> On the other hand having too many branches seems like the same kind 
> of "trap" that we got into with arch - defining branches and then no 
> one shows up to maintain them and no one bothers to merge them etc. 
> Maybe we should just sync the Silver main branch.

Hmmm, I would say that what now is axiom--test--1 should just become

silver/branches/axiom--test--1

or whatever name you like to give it. And it should live on Sourceforge
and on the wiki with identical data. The Wiki could be freely updated by
everyone (maybe some registration would be useful, so we would know whom
to contact if we like or dislike a modification). Then we (ax-dev
mailing list) have about one (or two?) month to give
opinions/testresults etc. If there is no serious problem then the
modifications can be merged with silver/trunk.

I strongly believe that axiom--main--1 should be a "testing" branch and
things from their should only enter the silver/trunk if they do not
break anything serious. Note that silver/trunk is supposed to become Gold.

On the other hand, patches to silver/trunk should be immediately applied
to silver/branches/axiom--test--1 and thus become available on the
website. axiom--test--1 is not supposed to compile, but it should be
close to it and should be at least as hot as silver/trunk.

> > Is that sync'ing easily achievable?

> Yes. We could automatically do an 'svn update' every evening. I have 
> been thinking about how best to view the online wiki version of the 
> source and documentation. I believe that it should be treated in much
>  the same way as one would treat one's own local copy of a
> repository. The only difference here is that this online version
> might be edited in a collaborative fashion. We should not think of it
> as "yet another repository" but rather just the online collaborative
> shared version of one of the main repositories. Using 'svn update'
> can be automatic but I think we might need to take more care with
> 'svn commit'.

I believe that people should register before they can change code. Then
commiting is not a problem, you just forward the registered name (and
allow it).

> Should we limit commits to only one file at a time?

I would say so, since we might accept some changes and reject others. It
is easier to pick the stuff if you have per-file-commits. I believe that
it is a bit hard to submit changesets through the webinterface. But it
should be required to add a reason for the change and that reason should
become an svn-log.

> Another alternative that I have been thinking about is only 
> supporting a "diff and patch" interface for the online wiki version 
> of the source and documentation.

Maybe some people like diff and patch, I don't. I am not a computer and
have problems in reading the diffs. It is simply not natural for a human.

> > I don't like very much that there is a separate (completely 
> > different looking branch book--main--1).

> I think originally Tim had in mind that there might be people who 
> only wished to work on the Axiom documentation.

Hmmm, that is strange. Axiom should become just one big documentation
(all is literate). Seems like the book stuff is a project that is
"about" Axiom but not Axiom itself similar to "The TeXbook" and
"TeX---The Program". If that was the intension? Then I could understand
the splitting, but I still believe that we should have everything in
just one repository. If I come to the Axiom project I would not like to
learn after some weeks of studying the sources that the Axiom-Book is
actually another project and doesn't live in axiom--main--1.

> > Reduce the number of different repositories!

> I agree! My personal opinion is still that svn is a poor choice: I 
> would rate it only slightly above cvs and below arch.

Yes, yes. And I also believed that arch is better than svn (and still
do), but Gaby is right in the sense that we must lower the barrier for
new people. That is more important than a non-perfect SCM.

> I still have a lot of trouble getting svn to work properly on
> windows.

That is not my main concern. I rather think that the missing starmerge
is a problem. SVK should overcome that, but I have not yet tested it
properly.

> But in the interests of easing cooperation and collaboration I think
> we need to compromise and standardize on svn.

Yes. Let's for the moment concentrate on svn/svk. If the SCM becomes a 
problem we can switch later.

> In fact I think we should drop the entire arch (tla) archive.

We can. It is as simple as this: remove any reference to any other 
repository than axiom--main--1 from 
http://wiki.axiom-developer.org/AxiomSources. If nobody knows about 
these archives, they are gone.

But we should ask Tim about it and, in fact, I believe, now that we have 
the AxiomSources page, they don't hurt that much. What I would actually 
like to see is how much alive a branch is. But maybe that is asking for 
too much. Is it possible to see branch activities on sourceforge without 
actually downloading the branch?

Ralf