We had hive developer's first meeting at Facebook on June 1.
I am sorry for sending the meeting notes late.
The following people were present:
Facebook (Paul Yang; Ning Zhang; Yongqiang He; Ahmed Aly; John Sichi; Ashish
Thusoo; Namit Jain)
Netflix (Eva Tse; Jerome Boulon)
Cloudera (Arvind Prabhakar; Vinithra Varadharajan; Carl Steinbach)
Yahoo (Alan Gates)
The following were the main meeting minutes:
1. We should have these meetings more often, say every month. Cloudera will
host the next meeting.
2. We should try to have a release every 4 months. We should try to push out
0.6 before end of June,
For the new release, Cloudera will take a lead on the release management
issues and also
help with documentation. Documentation for Hive leaves a lot to be desired.
3. The test framework is pretty brittle, and it is pretty difficult for new
people to do big
contributions without having a very sound test-plan. Ideally, facebook
should host a test
cluster so that everyone can run tests there.
4. A lot of external customers are asking for ODBC/JDBC support on top of Hive.
Cloudera will
take the lead on that.
5. The process of making a new committer should be more transparent. In order
to grow the community, it would be very desirable to
add more committers outside Facebook.
6. Create new components for Drivers (ODBC/JDBC) and UDFs.
7. Yahoo will take the lead of making Hive work on top of Zebra
Some new tasks were identified, but they can change if new priorities come in.
8. Carl will focusing on 'having' support and co-related sub-queries.
9. Arvind will be focusing on the cost-based optimizer
The main idea was that we should meet more often and share our ideas.
Time-based release will
be very desirable.
--
Thanks,
-namit
