This sounds good to me. The main point of critique I can think of is
that so far we haven't be able to do releases very fast. So in that
sense, the time schedule is probably very unrealistic. However, there
is nothing I would like more then us to be able to actually *do*
releases fast, so if this is another carrot on a stick to make *that*
happen, I'm all for it.
Eelco
On 3/7/07, Igor Vaynberg <[EMAIL PROTECTED]> wrote:
pasted from almaw's email on @user
-igor
-------------------------- 8>< --------------------------------------------
In my opinion we could, within the next:
-----------------------------------------
1 week - Push 1.3-betas as-is.
2/3 weeks - Bug fix as people test it and push out rc's when
we feel it's solid and stable.
4 weeks - Rename 1.x branch to 1.3.x.
- Release 1.3.0 final and put 1.3.x immediately into
maintenance mode.
- Create 1.4.x branch from 1.3.0 tag.
- Merge the model changes from trunk to 1.4.x.
- Backport anything else from trunk to 1.4.x that's
not JDK5-specific.
6 weeks - Push out 1.4-betas
7/8 weeks - Push out 1.4-rc's
9 weeks - Push out 1.4.0 final
- Create 1.5.x branch from 1.4.0 tag.
- Backport/add generics, covariance and other JDK 5 trunk
features to the 1.5.x branch.
- Move trunk to "2.0_deprecated_-_use_1.5.x_instead"
14+ weeks - Release 1.5.0
Suggestions to make this work:
------------------------------
We won't backport from 1.4.x -> 1.3.x.
We won't actively develop trunk.
We will push 1.4 out very soon after 1.3, and encourage migration.
We will have this in a public roadmap so people can see it coming.
Notes on what you think is insanity, but actually isn't:
--------------------------------------------------------
We will of course end up with five(!) branches (1.2.x, 1.3.x, 1.4.x,
1.5.x and what's currently trunk). This may seem like madness to you,
but I reckon it isn't:
During 1.3 development, 2.x is low activity, 1.2.x negligible.
During 1.4 development, 1.3.x and 2.x are low, 1.2.x negligible.
During 1.5 development, only 1.4.x will also be quite active.
Once 1.5.0 is out, we can properly deprecate 2.0. People currently using
it may not like being told to migrate to 1.5.x, but that shouldn't be
too hard (much less hard than going from 1.3->2.0) and there shouldn't
be too many of them. I guess that's the price you sometimes pay for
using unreleased software. :-/
I'd envisage 1.4.x will require some backports from 1.5.x. We'd
obviously encourage core developers and patchers to upgrade their sites
to use 1.5.x, do active development on that, and therefore try to only
ever backport from 1.5.x to 1.4.x, not forward-port the other way around.
If you think I'm smoking crack, the above is utterly unreasonable, you
want to kick me out of the gang, or you have any better ideas or
suggestions as to how to keep everyone happy, please shout now. :-)
Best regards,
Alastair
