🚀 go-pugleaf

> I'm investigating version control systems for our medium sized project.  I've 
> used ancient versions of PVCS and MKS RCS.  I have a recent Test Drive version 
> of MKS's Source Integrity.

I've used both PVCS and MKS RCS. PVCS, at the time I started
working in NT, did not support long filenames. This may have
changed. I would have used it but for this limitation, because it
was always highly reliable before (DOS and Win 3.1). PVCS has
the advantage over RCS that you can integrate it with a make
system which will check out the file if the source modification date
is newer than the object date (if you set up this dependency for
RCS, it will check out all files which have been touched, but
haven't had the source modified, such as when you add someone
to the access list).

MKS RCS for NT has been a major disappointment. We're struggling
with a problem right now -- it won't correctly check out branched
versions of files. This is _fatal_ for our project. The other major
problem (of which MKS has been aware for over a a year now,
and still hasn't fixed: if you use the 'co -g' version of checkout,
it's supposed to preserve the file modification time; however,
it always gives you the time in Greenwich meridian time if your
time zone is set for anything other than Greenwich. If you set 
your time zone to Greenwich, the time on checkouts is correct,
but the times listed with MKS's version of ls are wrong.  It's a
problem because you then have to use co without the -g option,
which gives everything a new date stamp, which forces everything
to rebuild (even if only one file has changed); but it's too dangerous
to use co -g, because it won't cause a file to be recompiled that
actually is newer than its object. I would _very_ strongly advise
that you _not_ use MKS RCS.

I've heard very nice things about SourceSafe, but haven't had
a chance to use it.