Thank you all for figuring out a solution for the security pull request.
Lets try to get 1.2 feature freezed as fast as possible so that we can "unblock" waiting features like FLIP-6 and the remaining security changes. *What do you think about Friday evening (6pm Berlin, 9am US west coast) for feature freezing Flink 1.2?* (only bugfixes are allowed in afterwards) I'll then fork-off a "release-1.2" branch and update the version in "master" to 1.3-SNAPSHOT. Please object if you have a bigger change or any other reservations regarding the feature freeze date! This is my current view of things on the release: - RESOLVED dynamic Scaling / Key Groups (FLINK-3755) - RESOLVED Add Rescalable Non-Partitioned State (FLINK-4379) - UNRESOLVED Add Flink 1.1 savepoint backwards compatability (FLINK-4797) - RESOLVED [Split for 1.3] Integrate Flink with Apache Mesos (FLINK-1984) - UNDER DISCUSSION Secure Data Access (FLINK-3930) - RESOLVED Queryable State (FLINK-3779) - RESOLVED Metrics in Webinterface (FLINK-4389) - RESOLVED Kafka 0.10 support (FLINK-4035) - RESOLVED Table API: Group Window Aggregates (FLINK-4691, FLIP-11) - RESOLVED Table API: Scalar Functions (FLINK-3097) Added by Stephan: - NON-BLOCKING [Pending PR] Provide support for asynchronous operations over streams (FLINK-4391) - NON-BLOCKING [beginning of next week] Unify Savepoints and Checkpoints (FLINK-4484) Added by Fabian: - ONGOING [Pending PR] Clean up the packages of the Table API (FLINK-4704) Move Row to flink-core ( Added by Max: - ONGOING [Pending PR] Change Akka configuration to allow accessing actors from different URLs (FLINK-2821) On Mon, Dec 12, 2016 at 5:43 PM, Stephan Ewen <se...@apache.org> wrote: > Hi Vijay! > > The workaround you suggest may be doable, but I am wondering how much that > helps, because the authorization feature would be incomplete like that and > thus of limited use. > > I would also assume that merging it properly and in full use after the 1.2 > release would be a bit better - in general, we have often avoided last > minute additions of sensitive and complex features. > > Do you think it is more urgent to have this in Flink? > > Best, > Stephan > > > On Mon, Dec 12, 2016 at 2:49 PM, Vijay <vijikar...@yahoo.com.invalid> > wrote: > > > Max and Ufuk, I respect your concerns and fully understand the importance > > of the network layer stack in Flink code base. Will you be comfortable to > > merge the code if I remove the Netty layer changes and leave the rest of > > the code. We can address the Netty code changes post 1.2 release? > > > > Regards, > > Vijay > > > > Sent from my iPhone > > > > > On Dec 12, 2016, at 3:38 AM, Ufuk Celebi <u...@apache.org> wrote: > > > > > > On 12 December 2016 at 12:30:31, Maximilian Michels (m...@apache.org) > > wrote: > > >>> It seems like we lack the resources for now to properly to take > > >> care > > >> of your pull request before the release. Unless someone from > > >> the > > >> community is really eager to help out here, I would be in favor > > >> of > > >> merging the pull request to the master after the release branch > > >> has > > >> been forked off. We should make sure it gets the attention it deserves > > >> then. > > > > > > Thanks Max! I fully agree with your reasoning. +1 to not include this > in > > 1.2 now, but look at it afterwards. I hope that OK with you Vijay. > > > > > > – Ufuk > > > > > > > > > > >
