[e-lang] Source control for E?

[David Hopwood]

Mark Miller wrote:
> On 5/12/06, Dean Tribble <tribble at e-dean.com> wrote:
> 
>>Based on everyone's input, I think MarkM is going to start with using
>>subversion on SourceForge.  Thanks for your time.
> 
> Uh, I hadn't intended to close off debate this early. I am leaning
> towards subversion, but I'm still reading about Monotone and Darcs.

It's not unexpected that the readership of this list would be likely to
know about or advocate SCS systems that embody interesting new ideas,
like darcs and Monotone.

I'd like to suggest that the temptation to use these newer systems
should be resisted for the E project, and that a more appropriate
choice is a safe, boring, known-quantity SCS that has had several years
to iron out bugs and to develop robust tool support (but that isn't
broken by design like CVS). Of the open-source alternatives suggested
so far, Subversion is the closest to this. Let boring application
software test out bleeding edge SCSes :-)


On the topic of darcs, I have used it a little for the Slate project
(slatelanguage.org). It was definitely easy to submit patches, and
*all else being equal*, I think changeset/patch-oriented SCSes like darcs
have significant advantages over revision-oriented SCSes. However, at
one point my local copy got into a state where I was unable to merge
changes, and had to check out a new local copy, manually merge across the
files I'd been working on, and delete the old copy. I am not sure that
this was a darcs bug; it may just have been pilot error. For the above-
mentioned "boring application software", I'd say risk it and go with
darcs -- this kind of corruption (if it was corruption) is possible
with any SCS, and is almost always recoverable. However, I don't think
that darcs is quite ready for use as the E project's SCS.

-- 
David Hopwood <david.nospam.hopwood at blueyonder.co.uk>