FANNG1 commented on PR #4232:
URL: https://github.com/apache/gravitino/pull/4232#issuecomment-2287698072

   > Hi @xiaozcy thanks a lot for your work. I think most of us are debating on 
how to make a meaningful property. Well, this indeed is a problem that should 
be carefully thought of, I was thinking of a more fundamental problem: how do 
we support different storage?
   > 
   > As we may continue to add more storage support, like adls, gcs, oss, 
adding all of them as dependencies for Hadoop catalog will make this catalog a 
dependency hell. In the meantime, for most of the companies, one or two 
storages should be enough.
   > 
   > So I was thinking of introducing a pluggable framework to support 
different cloud storages, users can configure to make the storage they want 
work, no need to package all of them together.
   > 
   > Can you guys please investigate a bit about other projects like Iceberg, 
etc, to see how they handle this problem and make an elegant framework? 
Basically, what I want is a pluggable framework for different storages, users 
can configure to make it work, besides, they can be in the separate packages, 
unless you make it work manually, the storages cannot be worked automatically.
   > 
   > Meanwhile, I think to support S3, we should also make it work in both Java 
and Python GVFS implementations, otherwise it cannot be used end to end.
   > 
   > So @xiaozcy @yuqi1129 @FANNG1 can you guys please think about my questions 
and propose a better solution?
   
   I agree that it's proper to provide a plugin method, but this seems not 
urgent and a blocking matter. Iceberg is a library which should not introduce 
too many dependencies to users, which Gravitino is a server which hidden the 
dependency issues to user and `Polaris` and `UnityCatalog` doesn't handle this 
too, IMO it's more importance to provide a unified storage configuration for 
all catalogs than providing plugin storage.


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to