Re: Branching 3.0?

On Sat, 2003-04-26 at 03:00, Henrik Nordstrom wrote:

> First there should be a feature freeze. When the feature freeze have
> cooled down and settled PRE1 is in order.

Sure.

> Making a PRE release before a settled feature freeze of the version is
> not in order. We did this in 2.5 and it was a big mistake which
> seriously delayed the 2.5 release and probably also impaired the
> quality of Squid-2.5.

Agreed.

> > Oh, and I do have further things in the works, but I'd rather not
> > be destabilising 3.0 after having tackled all the reported bugs...
>
> My opinion on when to enter feature freeze: The feature freeze should
> only be entered if it can be estimated that it really is a feature
> freeze and that the feature freeze for the next following version is
> at least 6-10 months away.

Well, I think that this is going to be difficult: All the developers
have various projects approaching 'integrate to HEAD' readiness. There
is -no- way that we can sensibly put everything aside for 6-10 months.
Thats why branching is so important: we feature freeze - and I mean
freeze - the branch, and -only- release blocker bugfixes get committed
(That is, bugs we would normally patch a released version to fix).

> Regarding Squid-3: The code base has undergone rather intrusive
> changes since the last STABLE (2.5), which is an indication that the
> amount of required testing and bug fixing is likely to be rather
> high. As a result of this it is extra important the code base is
> properly feature freezed while testing. The changes was already large
> before the switch to C++ and refactoring, and has only grown larger.

Right. I couldn't agree more. This is why I've been pushing to get
squid-3 feature frozen and branched.

So: For clarity:

I have many changes that are pipelined for HEAD in the not-to-distant
future.
I don't want any of them in 3.0 (with one exception (a bugfix for ESI
that I'll commit tomorrow)).
I'm very happy to hold off merging stuff to HEAD until we branch, by
mutual agreement, but:
  * Getting *some* version of squid-3.0 in front of experimenters is the
fastest way to get large scale feedback. Adrian has been running
polygraphs, so (I believe) has Joe.
What label shall we give the feature frozen squid-3 so that
experimenters will do so?

Rob

-- 
GPG key available at: <http://users.bigpond.net.au/robertc/keys.txt>.