[Top][All Lists]

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

Re: [Axiom-developer] RE: GNU Arch - was patches

From: root
Subject: Re: [Axiom-developer] RE: GNU Arch - was patches
Date: Wed, 29 Mar 2006 20:05:49 -0500


> I want to say that tla/GNU arch is not at all looking good as a source
> control system (it's problematic on Windows and looks like it is barely
> alive as an on-going project) and I would suggest dropping it as a tool
> for the Axiom project.

I'm currently working on projects that use CVS (doyen), SVN (magnus/work),
Arch (axiom), and git (personal stuff). I don't like any of the systems.
All of them seem baroque and broken. I do admit that arch is a challenge
to use.

Complaining, however, is different from advocating.
What exactly do you advocate and why?

> I am perturbed by the splitting of the Axiom effort into separate Arch
> source trees/sub-projects which seem not to be able to remerge (due
> apparently to lack of developer time and/or interest) and which are
> presumably becoming more divergent from the core CVS source repository
> due to ongoing bitrot.

Except for the control files (.arch vs .cvs) the code repository on,, and are all
bit-for-bit identical. Every change I push to one of those I push to
all of those. Where do you see bitrot?

> I realise that you were writing with respect to a more specific issue,
> but your recommendation prompted me to write in respect of the more
> general issue of the Axiom source code control system.
> For that reason I changed the topic of the thread.
> I further say that the multiplicity of unresolved source control systems
> and sub-projects in the Axiom project is just one sign of what is
> looking more and more like an overgrown and somewhat choked garden to
> me.

The only "subprojects" in the axiom project are specific branches that
have been created to separate various developers who planned to work
on separate threads. These do not impact the main branch in any way
until the work is complete and I merge the two branches. For example,
Bill Page has a windows branch.

What developers do with their separate subprojects is entirely up to
them. I have no opinion on how/when/if they maintain them until the
time comes to merge the systems.

All changes and merges to the "golden sources" on the main branch at
the 3 primary sites are done by hand, hand checked, and hand tested so
it matters not at all which repository system we use. I don't trust
the golden sources to automated tools. For me, it's a matter of quality.

> I'm actually rather a fan of overgrown gardens, but I do like to feel
> that I can walk through them to get at the flowers without falling into
> a well.

The current local count (on my disks and various machines) shows that I
have upwards of 20 different branches working on separate problems like
the mac port, the FC5 port, the ansi port, the X-on-windows effort,
the provisos effort, the maple-integration-bugs effort, the next book, etc.

While it's a rather overgrown garden (more like a swamp) none of these
branches are visible to the outside world.

I make every effort to make sure that the code that appears in all
of the repositories is clean, tested, round-trip retested, and is
the latest version available at the time of the update. They are
all identical by design.

If you find that is NOT the case please bring it to my attention
and I'll fix it and the process that caused it.


reply via email to

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