Re: Crazy idea - replace RCS backend with ClearCase...!!!

[Julian Gosnell]

James Melton wrote:

> The only way your idea works is if the CVS server is carefully
> architected to separate the storage mechanism from the rest of the code.

Because I understood that cvs actually called the rcs binaries (or did last
time I built it) I reasoned that this was probably the case. Then I heard
that cvs now calls rcs code instead, so I reasoned they were consuming a
library produced as part of the rcs package, so this library now fulfilled
the same function. After a little bit of looking though sources and playing
with rpm I have now almost convinced my self that they have cut-n-pasted
all the necessary code from rcs into cvs and cut themselves loose ! Who
says code reuse is dead ! If this is the case (I invite anyone who knows to
put me straight here - please!) then I am just about ready to give up on
this idea, nice though I thought it was.


>
> I don't know if it is, and apparently no one answering on this list does
> either, which is itself a bad sign.

So what should I do - post the development list? The blurb on the website
(cvshome) suggests that this is used mainly for cvs interactions and that
discussion takes part on this list.


>
>
> Are you looking for dual interface capabilities, i.e. People can
> checkout/in, label, branch, etc. using ClearCase and CVS concurrently?

That would be useful, but initially just being able to talk CVS from
anywhere and have you code transparently kept in a vob would be nice


>
> The closest thing to a reasonable sounding approach to doing this came
> from Noel L Yap:
> > It might be easier instead to create scripts that look like CVS,
> > but calls CC underneath.

This was my first idea, then I decided to make my life more difficult.

>
> However, if you are also hoping to set up a remote CVS server this won't
> work - you need all the network code from CVS.

Plus, SourceForge (one of the pieces of s/w I am interested in) as well as
many others, looks for and parses CVS implementation files in the CVS
repository. If I continue to use CVS all these files may stay the same,
otherwise I would have to reengineer them or every piece of s/w that needs
them.

>
>
> If you don't need update access from both CVS and CC, I would create two
> repositories where one is a read-only replica of the other. Automating

> replication in one direction sounds much easier than what you describe.

>

I would prefer a solution where ultimately I would be able to work from
both sides.

This suggestion is my emergency parachute. I'm just investigating other
possibilities before I deploy it !


Thanks for all the help folks.

Please mail me any more thoughts/ideas.

I shall keep you posted.


Jules



>
> Good Luck and keep us posted,
> Jim Melton.


_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com