Re: [Gnu-arch-users] "virtual" archives in arch

[Mirian Crzig Lennox]

address@hidden (Tom Lord) writes:
>
> Other people have approached the same problem in a slightly different
> way -- in fact that's the origin of some of the CVS tools.   You might
> find it simpler to set up a "gateway" that propogates changes back and
> forth between a separate arch archive and CVS repository, handling
> just the special case patch flows you need.    There's little to be
> gained and probably insurmountable amounts of complexity involved in
> trying to map arch directly onto CVS (and vice versa).

Perhaps I'm being dense (it is a common occurrence), but I don't see
how the gateway idea solves any of the problems people are bringing
up, especially the crucial one of atomicity.  The whole point of
having Arch deal directly with the CVS pserver is that the pserver (if
it is the sole means of access to a CVS repository) provides the
atomicity that gives this idea any prayer of working at all.

To get anywhere, there needs to be a mechanism for locking out Arch
commits whilst a CVS commit is in progress, and vice versa.  This
problem no less exists where interaction happens through an
intermediary than directly.  Additionally, what appeals to me about
doing it all inside Arch is that it has this lovely internal interface
for describing archives at a higher level of abstraction than could be
possible as an external gateway.

cheers,
Mirian