Re: [9fans] revision control

[Paul Lalonde <plalonde@telus.net>]

On 23-Jan-06, at 3:18 PM, Ronald G Minnich wrote:
> These are wonderful things, don't get me wrong, but they're not  
> really good enough to be an SCM.
  Agreed.

> CVS is ok.
Barely.  I can't believe how difficult it is to put together a good  
revision control system.
The last couple of months were an interesting recapitulation of  
revision control for me.  Started a new job, at a startup, without an  
SCM.  CVS it was, and within the first branch/integrate cycle we were  
using Subversion.  A month later we've dropped coin on Perforce just  
for it's ability to manage branches and integrations in a sane way -  
something I haven't seen in any of the open source offerings.

The key problem is that most of the open source seem to support the  
open source development model: one person does all the integration  
into the main branch, usually re-writing the bulk of the change on  
the way through.  Integration into dev branches is rare, done only  
(it seems) at releases.  Heck, dev branches seem rare, and frequently  
seem to  be interpreted as forks rather than concurrent efforts.

Instead, at least in my work, dev branches are common: I run 2-3 at a  
time, corresponding to different bugs/features that I'm working on at  
the moment.  Integration into the main tree happens 2-3 times a week,  
integrations from the main tree daily.  The other developers do the  
same - integration *has* to work smoothly in that kind of environment.

If only subversion could understand what changes had and hadn't been  
applied to various branches.

Does anyone know an alternative to Perforce (which doesn't work so  
well disconnected) that does branching and integration so easily?

Paul