[GitHub] [flink] zentol commented on issue #8831: [FLINK-12935][hive] package flink-connector-hive into flink distribution for HiveCatalog

[GitBox <gi...@apache.org>]

zentol commented on issue #8831: [FLINK-12935][hive] package flink-connector-hive into flink distribution for HiveCatalog
URL: https://github.com/apache/flink/pull/8831#issuecomment-508742422
 
 
   Can we leave it out for 1.9 and revisit this for 1.10? There are some ideas floating around for splitting the repository (I believe Stephan wants to open a ML thread soonish), and this may or may not have repercussions on what is bundled in flink-dist.
   If this results in separate release cycles for connectors and the flink core (TBD), there may be no real point in bundling anything in /opt since it's likely out-dated before the next release of the connector. In this model, copying jars into dist would be a common upgrade pattern.
   
   I'd rather not add a new jar to /opt in one release if there is a possibility that we remove it in the next one again.
   
   This is admittedly an unfair argument given that this wasn't really brought up so far anywhere and there's lots of uncertainty.
   
   As for the proposal to add a `catalog` sub-directory, I had a similar discussion when it came to the plugin directory recently. My take was that it should have sub-directories for different categories, like `reporters`, `filesystems` etc., but it was rightfully pointed out to me that a given jar may fit into multiple categories.
   I'm wondering whether the same logic doesn't apply here as well; ideally we'd follow the same logic for `/opt` and `/plugins`.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services