RE: renames under CVS

[Greg A. Woods]

[ On Wednesday, February 27, 2002 at 06:27:48 (-0800), Noel Yap wrote: ]
> Subject: RE: renames under CVS
>
> So now you're saying that it's OK to rename the
> archive file under certain situations?

Yes.  Normally I would never suggest anyone manipulate the repository
directly, but there are always situations where the rules need to be
broken, and provided you know what you're doing, and you take all the
right precautions, this is one of those situations where no harm will
come.  You merely need to think about the consequences logically and
consider the past and future timelines of your project to see that in
the long term such a rename will merely become an invisible correction
to a minor aberation in a detail that's nearly invisible to future
onlookers, and it certainly will not affect the reproducibility of any
important milestone in your project.

>  Wouldn't this
> break the build for those who need to checkout by
> timestamp?

Possibly, but unlikely.  (in my test example the renamed file was
checked out with an old timestamp, but it in the case where there were
no local changes it was the same as the timestamp on the working file
under the original name, and in the case where there were local changes
my manual merge would obviously modify the file so make it newer than
any product derived from it)

> CVS is the _Concurrent_ Versioning System.  Hasn't
> your stance always been, "One shouldn't be striving
> for any degree of serialisation when using CVS"?

Indeed -- but renames are special and hopefully rare events, just like
feature freezes for release branching, etc.  You cannot avoid some
degree of serialisation when working with other people.  The goal of CVS
is not to eliminate all serialisation, only that encountered due to
edits in the compile-edit-test cycle.

> I know you've asked before, "Why are so many people
> asking for this feature?"  Have you ever asked
> yourself, "Why am I the only one defiantly opposed to
> such a feature?"?

I've never asked the question because the answer is obvious:  People
ignorant of the CVS implementation and inexperienced with its use will
ask for this feature because they cannot intuitively see that it's
neither really necessary, nor is it possible to implement in CVS without
making CVS into something else and losing certain very valuable features
it now has.  Better documentation may help reduce the frequence of the
question, and I've posted some suggested changes to the manual.

(Nobody's stopping y'all from implementing that new thing that is not
CVS but which works like it and which supports renames -- oddly though
nobody's implemented it yet in over 8 years of people complaining about
the missing feature! ;-)

[[ It really would help if your broken mailer wouldn't re-wrap
paragraphs incorrectly -- i.e. without preserving the attribution
indicators! ]]

> > > > Finally of course we'd need to find some way to assess the difference
> > > > between an ideal filename, and one that's suitable and sufficient for
> > > > the purpose.  While we might all like to rename many of the files and
> > > > directories in our projects, there's no fundamental necessity requiring
> > > > us to rename files that are not "grossly" mis-named.
> > > 
> > > This measure would be extremely subjective.
> > 
> > That's partly my point!  ;-)
> 
> Then it's not a metric.

Perhaps not -- which is why I was trying to make that point!

-- 
                                                                Greg A. Woods

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