Re: [boost] A *smaller* release concept

From: David Abrahams (dave_at_[hidden])
Date: 2007-10-17 08:09:35

on Wed Oct 17 2007, "Peter Dimov" <pdimov-AT-pdimov.com> wrote:

> John Maddock wrote:
>> Nicola Musatti wrote:
>>> I expect it to be much worse than that: trunk evolved for about a
>>> year and a half since the 1.34 branch was created. I don't think it's
>>> realistic to move new libraries onto 1.34.1 and expect them to work
>>> straight away.
>>>
>>> What I believe it's going to happen, judging from recent postings by
>>> some of the moderators, is that the new release branch is based on
>>> 1.34.1 and libraries are going to be moved over one by one in a
>>> controlled way, with insufficiently stable ones reverted to their
>>> 1.34.1 version (or just removed if they weren't in the previous
>>> release). This makes more sense to me.
>>
>> Correct, and anything not ready won't be merged to the new release at
>> all: basically the default assumption is "Ship 1.34", and folks have
>> to demonstrate that their new code in Trunk is ready to be merged to
>> release. But ultimately as release manager Beman has the final say on
>> everything!
>
> Yep.
>
> This is setting us up for a fork. The trunk will drift further and further 
> away from the release branch; if it's too hard now to reconcile trunk and 
> 1.34, it will be even harder to do that for trunk and 1.35, and nearly 
> impossible for trunk (if it still exists at all) and 1.36. At some point one 
> of the two branches will be cut off and someone will suffer, no matter which 
> branch prevails. 

Interesting you should say that;  Among the moderators who discussed
the 1.35 release process, all but one of us was very concerned about
that issue.  Another problem with the drift is that the trunk very
quickly becomes an invalid proving ground for a library's ability to
work in context of the release branch.

-- 
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com