arch, was Re: SourceForge has drifted away

MJ Ray markj at
Fri Aug 16 11:28:43 UTC 2002

Werner Koch <wk at> wrote:
> I have only looked at the source and frankly it reminds me somehow on
> Dan Bernstein's code (except that Tom is clueful about Free Software
> and he has comments in his code).

Interesting.  I don't really get that impression, but then again, I haven't
had that much exposure to Bernstein's code.  I'm not willing to praise the
code to the highest, though... more sort of withhold judgement for now ;-)

> This is a misunderstanding.  I mean the build system; it is not
> autoconf/automake based so it is hard to start working on arch it.

As someone who's failed to grok autoconf (I should really retry now -- it's
been nearly a year), I can see his reasons, even if it makes life harder for
other developers.  I don't know whether it would be possible to add
autoconf: as someone who knows autoconf, you do have a snap opinion on that?

> Having common standards to build software is a good thing becuase it
> saves a lot of time to learn the boring details on how to build
> something.

Actually, the arch formats and algorithms are being documented with the aim
being to create other competing implementations.  Having common standards to
software formats is a good thing too, isn't it?  Hopefully at least one of
the competitors will use autoconf et al.

>> If not, where there is a performance benefit, some parts have been converted
>> to C, I think, which may be where the C library came from...
> Shell skripts are in general not evil but the sheer amount of them in
> arch makes me worry and they definitely don't help in debugging.

Small modular parts are helpful in debugging.  Shell script is probably
better than some of the alternatives at this stage.  It's not fatally wrong
and it may even be optimal for some parts that just launch other programs.

The problem I've noticed with arch's code are the couple of places where
there are breaks in the modularity and undocumented incestuous relationships
between parts.  That's not good.  It's acknowledged as a problem, but
unfixed so far.

In summary: I think this project is interesting and I'm now blundering
around, trying to find out more about it.  How it compares to the other
source control systems, I'm not yet sure.  It seems nearer to the Aegis end
of my scale than the CVS end, but I've not yet seen whether it can help with
development policies.


More information about the Discussion mailing list