On Wed, 2002-10-09 at 01:10, Henrik Nordstrom wrote:
> Robert Collins wrote:
>
> > I understand the difference you are making between the two ways of
> > expressing it.
> >
> > 2.6: as per the TODO.
> > 2.7: nothing definately slated. (Some possibles in the TODO).
> > 3.0: ????
> >
> > So, when you say 'begin development on 3.0', I'm not sure how you
> > differentiate that from 2.6/2.7. Other than language that is. And if
> > language is the only differentiator, then language IS the metric for
> > having 3.0 ready.
>
> The 3.0.STABLE is in my opinion at a minimum the version after language
> conversion plus refactoring needed to make us reasonably happy about the
> code.
I really think we should separate the two things. The code we have today
doesn't change much to become C++. Making us reasonably happy across the
board is a 100,000 LOC review and refactor job. Starting a process of
continual review will be easier and more long term effective IMO. That
means refactoring an area of code before we add features, no more, no
less.
> If we start 3.0 now the likelyhood for a 2.7 release decreases significantly
> as most new development should then be done in the 3.0 branch where possible,
> and the stability of 2.6 is also likely to suffer due to less attention by
> developers.
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.)
> > Also, from a risk management perspective, we should release a C++
> > version with minimal C++ features as soon as we complete the conversion,
> > to get feedback and provide a baseline.
>
> Which may well be 3.0.DEVEL1 in my opinion, or what users do you plan on
> target with this C++ version?
All our normal users. If we are happy that it is as good as 2.5S1 or
better, then we can release it.
Rob
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:16:54 MST