> I don't think what we currently do is really the best. I 
> think your notion of having a branch where work is done for 
> the impending release probably makes more sense. Leaving HEAD 
> for the most current development processes target for 
> subsequent releases also makes more sense to me as well.

I agree. We could do this now by declaring the RC2 branch as the 1.0 branch,
and making HEAD 1.1.
(when it comes down to details, I'd prefer to actually recreate the branch
so it's a cleaner merge as it currently contains parts of HEAD up to the
last merge tag point, but we can discuss this later).

> But I agree with Dion that you cannot start making changes or 
> additions to HEAD which is currently slated for 1.0 until we 
> figure out what we're going to do wrt what branches will be 
> for what versions and what HEAD actually represents to us.

The main problem with using HEAD for "unstable" development is that people
start to complain when it doesn't work properly. I think we'll get over that
as we have separated the plugins and move towards better tested, component
oriented development as it is less likely to have the same severe impact as
ripping apart the most central class has had this time around :)

> 
> Vincent, I'm all for what you suggest as it means we can 
> actually develop easier in parallel but some simple naming 
> schemes and a small paragraph describing how we move from 
> branch to branch would be nice.
> 

I agree. For a short description, you can read the section "the flying fish
approach" in the CVS Open Source Development book:
http://cvsbook.red-bean.com/cvsbook.html#The_Flying_Fish_Approach_--_A_Simpl
er_Way_To_Do_It
(warning - this is a 600K HTML document for those on a modem - here is the
links to other downloads http://cvsbook.red-bean.com/ :)
The whole chapter on branches is quite good.

Cheers,
Brett

Reply via email to