RE: cvswrappers - any better suggestions ?

[Greg A. Woods]

[ On Monday, April 2, 2001 at 11:22:32 (-0700), Paul Sander wrote: ]
> Subject: RE: cvswrappers - any better suggestions ?
>
> Greg's statement below is flatly untrue.  We've discussed this very topic
> at length in this forum many times in the past.  His argument is based on
> the fact that the RCS "merge" tool cannot support merges of arbitrary
> file formats.

Well, sort of.  It's a little deeper than that, of course.

>  It doesn't even support merges of arbitrary ASCII formats.

Well, no, but that's partly the point!  ;-)

> And yet, the CVS community attempts to use it in that capacity all the
> time.

And with remarkable success, success that's been scientifically
documented ever since CVS-II was released.

Indeed there are several other tools with the same basic design for
support of concurrent editing and they are equally successful.

Of course programmers who have used CVS for a long time will inherently
learn little tricks that make actual merge conflicts even less likely to
occur except where absoutely necessary.  Indeed there are even
programming tricks that can make accidental hidden conflicts less
likely.

And of course the CVS source itself provides a half-decent example of
the benefits of automated product testing.

> Inserting a registrar into CVS to allow shops and users to specify the
> particular tool required to perform a merge is not a fundamental change to
> the CVS design, but it is a small generalization.

Yes, but one that would require a rather drastic revision to the
repository format.

>  And it's one that will
> greatly benefit the CVS community in general.

I'm not so sure.  If it were so it would have been done long ago and
offered back to the community.  As it is we don't even have a sample
implementation to show to the community and to prove that such a forward
change in the repository format would succeed.

>  Such a mechanism does NOT
> affect in the slightest way ANY aspect of the concurrent development model
> that CVS implements.  (It is basically just parameterizing the hard-coded
> path to the RCS-supplied merge tool, or its equivalent in the librified RCS.)

I agree whole-heartedly.

I'm just not restricting my view to one set of concerns....

-- 
                                                        Greg A. Woods

+1 416 218-0098      VE3TCP      <address@hidden>      <robohack!woods>
Planix, Inc. <address@hidden>; Secrets of the Weird <address@hidden>