[asterisk-dev] Release schedule ?

[John Lange]

On Tue, 2006-11-14 at 16:48 -0600, Kevin P. Fleming wrote:
> Russell Bryant wrote:

> I think the biggest problem we had is that we committed to too many
> features going into this release, before they were actually ready to be
> merged.

Ironically the pressure to get stuff in is actually going to mean it'll
be longer before we can benefit from the new features because all that
new stuff is going to be very hard to stabilize.

> In other words, either we continue adding this much new functionality
> between major releases and live with the fact that it will take 9-12
> months to make releases, or we bite the bullet and make the decision to
> start the release process even without functionality we wish to have in
> place.

I believe one is a function of the other. If major release increments
were on a shorter cycle then there would be less pressure to force
something in before its ready because it won't be long before it makes
it into the next version.

On the other hand, if its a high demand feature and you know the next
version is 2 years away there is a lot of pressure to force it in.

Would it be too radical to identify the top 1 or 2 significant features
along with any minor ones and roll out a new significant version every 6
months? The paid development staff would be able to show a road-map at
least 2 or 3 versions ahead. Nothing saying it can't change over time as
priorities change but it shows strong direction.

Or does that kind of quick release cycle just make people angry because
they have to upgrade too often or risk falling woefully behind? One way
to mitigate that is to increase the backwards compatibility time to
major revision number increments rather than significant increments. (In
other words, all 2.X versions would be compatible rather than 2.1.X like
it is now).

John