Re: C++ (again) / Squid 3

On Wed, 2002-10-09 at 15:56, Henrik Nordström wrote:
> On 9 Oct 2002, Robert Collins wrote:
>
> > Ah - you are suggesting a 3.0 branch. I'm against this. Every time I've
> > seen a -major- change done outside of HEAD on a project, it suffers
> > badly. I'd like to see HEAD become C++, as a single series of commits.
> > Theres no need for a branch for that. (the rbcollins_cxxtest on sf could
> > be such a branch, although I've been using it to test the differences
> > that C++ allows us to make, as opposed to a straight syntax conversion.)
>
> Yes, I am suggesting a fork. Squid-2 for continued management of the
> current C codebase, and HEAD for C++ transition with the goal of a Squid-3
> release.

Hmm, I'm not to worried what we call it version wise 3.0 is fine to me.
As long as it happens in HEAD, I'm happy. What you describe above seems
fine to me.

> > All our normal users. If we are happy that it is as good as 2.5S1 or
> > better, then we can release it.
>
> I am not too happy about making a STABLE release that does not add any
> significant functionality and that know we will immediately refactor large
> parts of the internals.

I've not suggested that we do a STABLE release with no extra
functionality. I've suggested that we don't go wholesale refactoring
until after the first C++ release, as a risk mitigator.

I.E.:
To avoid doing a /null/ STABLE release, we could simply drop the C++
patch in with (say) 2.6 (and call it 3.0 if we need to communicate the
difference to the users). Giving the user significant extra
functionality (rproxy, ETag, ESI, IP6, windows) (picking a few from the
TODO), and giving us a C++ baseline from which to build on.

Rob