Re: [Gnu-arch-users] DARCS

[Mirian Crzig Lennox]

On Fri, 5 Sep 2003 16:06:17 -0700, Zack Brown <address@hidden> wrote:
>
> Darcs is really cool. One thing you don't mention in your post is that
> darcs supports arbitrary characters in filenames, which arch doesn't do
> (and apparently will not). That's a big deal.

I do like that.  Darcs seems to be especially well suited for use by a
package maintainer or an occasional bugfixer, where one has little or
no control over the organisation of the code.  Much as I like Arch, it
is quite intrusive, much more so than I would prefer.  My impression
is, however, that the Arch community in general does not see it as a
problem.

> David Roundy (the darcs maintainer) is very nice and doesn't chew people
> out and belittle them the way Tom does. That could be considered a
> feature.

I agree, to a point.  Every successful open-source software project
I've ever known has had a leader with a strong personality and sense
of vision, who isn't afraid of stepping on some toes now and then.  It
seems to be crucial for keeping members committed, and the effort from
splitting off in too many different directions. 

However, I do not entirely share Mr Lord's opinion that confrontation
is a productive means of collaboration; I think that confrontation,
while sometimes unavoidable, tends to alienate contributors who are
productive but timid.

On the other hand, I have no experience with David Roundy's style of
leadership.  In fact, I have been trying to contact him, but
abridgegame.org has been dead for several days running.

> The learning curve for darcs is very gentle. You can basically start
> using it right away, without any agony at all.

Yes and no.  Arch's documentation, which I read before I started
playing with Arch, is really quite good.  Whilst it was easy to get
myself started in Darcs, it also took a lot more trial-and-error
figuring out how to do the equivalent of Arch's "hello world"
tutorial.  Arch also has exemplary online help; Darcs' is but
passable.

> On the flip side, darcs is a much younger project with a much
> smaller following and is still struggling with some fundamental
> design issues like how to distribute a repository. Being written in
> Haskell, darcs will inevitably have performance issues (although
> David feels these will be minimal), as well as trouble recruiting
> developers. It also seems to have issues with conflict resolution,
> that may take awhile to really resolve.
> I think darcs could turn into a really nice tool, but not before arch makes
> the headlines. Arch has made a really startling amount of progress lately,
> or maybe it only seems that way because up till recently it was "just"
> a bunch of shell scripts; and that made it hard for many people to take
> seriously. Whatever the case, it is looking like arch is now usable, and
> about to become very popular. Darcs just isn't at that stage yet.

To me, Darcs has more of a feel of a research paper and accompanying
reference implementation than a production-quality piece of software.
This is not necessarily a bad thing, however.  The advantage is that
we have a very rigorous formal specification and theory to go by, as
well as a working prototype.  The next step might well be for someone
(perhaps an expert coder rather than a theoretician), to start on a
next generation implementation in C or C++.

As for Arch's popularity, I think it's far too early to form a
definite opinion.  CVS remains the sixty stone gorilla of the
open-source revision control world, and we've a long way to go before
that ever changes.

cheers,
--Mirian