2.1 / 3.0 roadmap

[Jonathan Ellis <jb...@familyellis.org>]

I'd like to focus 2.1 on technical debt as well as performance.  I'm
okay with not having big features to announce, other than custom types
for CQL.  If we focus on these, we should be able to hit a 2.1 release
in January, with 3.0 back on a six month schedule for July.  I note
that we're still catching bugs introduced in 1.2.0; a short release
focused on plumbing over features should give us a chance to catch up
and release a 2.1.0 that can be more stable rather than less.

We had a separate thread on performance tickets, summarized at [1].  I
think these are the critical technical debts we need to pay off:

- Counters (CASSANDRA-4775)
- Repair overhead (CASSANDRA-5351 and possibly 3362)
- Making CFIDs globally *and* temporally unique to prevent "re-using"
data from an old incarnation in a drop/recreate cycle (CASSANDRA-5202)

DataStax has engineers working on these, so I'm not asking for
volunteers, but if there are other improvements that should be tier-1,
I'd like to hear about those.  (And volunteering would be great!)

I don't yet have firm plans for 3.0, but I did want to float the idea
early that we should kill CQL2 (CASSANDRA-5918).  By July 2014 it will
be 18 months after the introduction of CQL3 which is ample time to
migrate.  (Killing Thrift is still off the table indefinitely.)

[1] https://issues.apache.org/jira/issues/?jql=project%20%3D%2012310865%20AND%20labels%20%3D%20performance%20and%20status%3Dopen