[asterisk-dev] Thoughts on Asterisk release management

[Philippe Sultan]

Russell,

the motivation behind the new release management proposal is to have
the trunk code actually tested by Asterisk users. I think we can
distinguish between target users depending on their profiles.

Shorter release cycles might make users tracking even small new
features in Asterisk test new releases. This group of users is really
essential, and unfortunately quite restricted.

Other people, who run small or medium Asterisk installations, will be
willing to test a new release if (in that order) :
- they really need a new compelling (set of) feature(s) on their
Asterisk server ;
- chances are good that their Asterisk server won't pop up a lot of
new bugs, especially if they look unrelated to the feature(s) they
wanted to test.
The first point is very subjective, and new features probably need to
be restricted in number as they cause new bugs. Therefore, I think the
latter point will definitely be improved by shortening release cycles,
and by following the policy you've exposed regarding the trunk
development.

Finally, I also agree with Mihai saying that stress tests are
essential to large Asterisk installations administrators. Many will
prefer stability over new features. Packaging testing tools along with
new releases might help them step towards latest versions of Asterisk.
I understand though that it requires a huge amount of work, which is
probably out of the scope of the immediate changes you're thinking of.

Philippe