Here are some brief notes I took during the hangout: Attendees: Chris, Jacques, Byron, Parth, Sudheesh, Daniel, Hakim, Jim
Jacques: release schedule for the rest of the summer proposes the 15th of each month for the next couple of months Chris: I'd like to move to 6 weeks It hasn't really been exactly 4 weeks yet anyway. Propose a strategy around bug tagging in a more open way. Until we have a strong commitment to do something, we should leave them as "Future." Constantly pushing bugs into a future release seems like a poor use of people's time. Parth: the problem with the future list, is it might just be a big dead bucket. Jacques asked about the state of the new buffer allocator... Chris: I'm almost done with some changes that eliminate locking in favor of transacted changes via a swap of an AtomicReference<> to an object that holds the state-tracking variables; this is an optimistic locking scheme that doesn't span allocators; transfers between allocators cause local "transactions" within each allocator. Jacques wants to get back to some old RPC tickets. Is the recycler worth using for the allocator? Jacques general rule: avoid objects for values within batches, but not for record batches themselves. Discussion of the use of the new serializing executor to deliver events to fragment executors. Hakim: we need to do some work with Parquet so we can catch up to them, and so they can take back our patches. Parth: we're moving forward with Calcite in order to be current on that as well. On Tue, Jul 7, 2015 at 10:02 AM, Chris Westin <[email protected]> wrote: > https://plus.google.com/hangouts/_/event/ci4rdiju8bv04a64efj5fedd0lc >
