begin  quoting Andrew Lentvorski as of Fri, Oct 20, 2006 at 01:58:20PM -0700:
> Stewart Stremler wrote:
> 
> >I think everyone does a little bit. But some groups do it more than
> >others; one of these days I should look at the CVS code "in anger",
> >just to see how hard it would be to add some features...
> >
> >(Versioning directories? Shouldn't be hard. Same for renames. Same
> >for offline/laptop use. I don't care about atomic commits, as 99% of
> >the time I commit one or two files, so that's not an itch for me,
> >but a global commit count might be interesting, especially if it can
> >be used an a $CommitNumber$ tag...)
> 
> My advice.  Don't.
> 
> A) Subversion does 90%+ of what you want

Subversion eliminates itself with it's dependencies, and then with
the attitude that comes along with 'em.  I won't play that game, and
"well, stop whining and download the binary then, pansy" is the sort
of attitude I do not care to be associated with.  (I also don't use
OO.o at home for the same reason.)

If perforce was open-source and as easy to compile as CVS, I still
wouldn't abandon CVS entirely.

> B) Snapshots are not quite the right abstraction
>
> Snapshots are a deceptively simple 85% solution.  And then you hit the 
> last 15% and realize that you can't get to the rest.

Um.... I've never even smelled that mythical 15%.

> Take a *long* look at the svn,

Eliminated due to dependency hell and developer attitude.

>                                mercurial,

Never tried. I should try it out.

>                                           arch,

I've successfully used it _once_, but I didn't quite grasp the underlying
abstractions.  It seems very clunky in practice.

>                                                 and darcs development 

IIRC, darcs depends on haskell. I had to compile haskell myself, which
wasn't a problem, but to compile it, I had to hack at the code in quite
unsafe ways. I don't trust haskell, so I can't trust anything built on
it.

> mailing lists before you consider this.  Especially arch(tla?), Tom 

MORE mailing lists?

I suppose I could give up the perforce mailing list. I don't read
most of the articles, on account of most of the contributors being
inveterate top-posters, rendering the digests incomprehensible.

> Lord(arch/tla) is far too histrionic, but his points have merit, and 
> there are generally links to the other lists where they discuss patch 
> theory.  The git, darcs, arch, and mercurial folks are all dancing 
> around some hard problems and trying to get them right.  It's not easy.

Noted.

I need to get tla up and running again anyway. I have GST 2.2b to
download and compile.

-- 
_ |\_
 \|


-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to