happy to share this week's community update including Flink 1.9.1, unaligned checkpoints, a new type inference system for UDFs in the Table API and a bit more. Enjoy.
* [releases] The vote for Flink 1.9.1 RC1 has timed out without a vote. I guess, Flink Forward Europe preparations in combination with Chinese national holiday... 
* [runtime] Arvid has published a FLIP to support unaligned checkpoints. Unaligned checkpoints could significantly reduce & stabilize checkpointing times in high backpressure and high load scenarios and result in lower latencies under checkpointing in general. 
* [runtime] Piotr has started a survey on dropping the "non credit-based flow control"-mode in Flink's network stack. No replies so far. So, if you are relying on it, now would be the time to raise your hand. 
* [logging] Gyula has started a discussion on adding contextual information (taskmanager, container, job id) to Flink's logs and is looking for feedback. 
* [sql] Timo has published FLIP-65, which proposes a new type inference mechanism for Table API user-defined functions. It is based on three levels of abstraction: automatic type inference, type hints via annotations and manual declaration. 
* [code-style] Zili proposes to drop "final" from method parameters under the assumption that all method parameters should be final anyway. Discussion ongoing. 
* Parag Kesar & Ben Liu of Pinterest have published a blog post on their Flink-based real-time experimentation platform. 
* Oran Hirsch of DynamicYield describes their data architecture including Apache Flink for stream processing in this recently published blogpost. 
* There will be full-day meetup with six talks in the Bangalore Kafka Group on the 2nd of November including at least three Flink talks by Timo Walter (Ververica), Shashank Agarwal (Razorpay) and Rasyid Hakim (GoJek). 
* Flink Forward Europe starts tomorrow in Berlin. Hope to see your there.