Anyone like us got comments on Perforce?

JasonTrenouthjason at harlequin.co.uk JasonTrenouthjason at harlequin.co.uk
Mon Apr 27 05:55:33 PDT 1998


On Sat, 25 Apr 1998 10:06:08 -0600, Wes Peters <softweyr at xmission.com> wrote:

> Robert D. Seals wrote:
> > 
> > At 02:42 PM 4/23/98 GMT, you wrote:
> > >company with multiple development sites around the world (ie vastly different
> > >time zones). These sites collaborate on many different projects and products.
> > 
> > Well, if they are connected at a reasonable speed, then p4 might work.
> > The problem is that p4 doesn't really support distributed depots; you can
> > reference an arbitrary number of depots, but only 1 of them can be
> > read-write, the rest all read-only. Again, this may be adequate, or
> > or it might not. If the sites are not well connected, then probably the
> > only real solution is to use several depots, and this is fairly hard
> > to manage, i.e., perforce doesn't do it, you have to.
> > 
> > So, if you can all use a single depot on 1 server, you're good with p4.
> > If the sites are connected (even slowly) and can live with read-only remote
> > depots, you're good with p4.
> > Otherwise, my perception is that p4 might not be ideal, and if anyone
> > knows better, let me know!
> 
> Having attempted to use several other SCM systems over dial-up lines, I
> appreciate the speed of P4 compared to, for instance, CVS pserver.  The
> atomic commits are also pretty important on somewhat unreliable lines.
> 
> As Robert pointed out, Perforce only really works with a central server,
> and you'll need reasonable connectivity between the remote offices and
> that central server.  The definition of 'reasonable' is going to vary
> depending on your expectations of speed.  For a single remote user, a
> 33.6 or 56K modem dial-up, or perhaps a dial-up ISP account and VPN into
> your network, is entirely adequate, and might even suffice for a small
> group of 2 - 4 engineers.  The next step up, ISDN BRI or ADSL, would
> probably support 10 to 12 users.
> 
> If you really need distributed source code stores, another solution may 
> be best for you.

Well, distributed stores are not a requirement for us, although they are used
as part the implementation of our in-house system.

Basically, we have multiple peer development sites in the US (x3) and the UK
(x4), with every prospect of more, plus a few people working from home.

We have reasonable connections between most sites now, with the chance that
they'll get better in the future.

The developers on a particular project are spread out almost randomly across
our sites, depending on where the skills are.

It sounds as if a reasonable Perforce-based solution for us would be to have
depots for our projects at sites that represented 'centers of gravity' for
those teams. As the team personnel changed over time it might be more efficient
to move a depot to the new 'center of gravity' for that project, but this would
be a relatively rare event.

This is sort of what happens at the moment in our in-house system except that
the granularity of 'mastery' (official copy) is at the level of a directory.

__Jason





More information about the perforce-user mailing list