Github user fhueske commented on the pull request:
https://github.com/apache/flink/pull/975#issuecomment-127405822
Hi @sachingoel0101,
allowing tasks to communicate with each other over side-connections breaks
the current assumptions that TMs exchange data only over data connections and
control communication is only happening between JM and TM. By allowing parallel
tasks to exchange data, these tasks are no longer independent. In fact the
system relies on these assumptions at several places including fault tolerance
& recovery. Even if in your use case, exact recovery might not be important,
other users might use this feature in a different way.
Changes that have the potential to change the behavior of the system like
this PR need to be discussed in detail by the Flink community. This can be done
by opening a JIRA and adding an initial design sketch or proposing such changes
on the mailing list and asking for detailed feedback.
I don't think we can merge this into Flink without breaking the system at
several places.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
