[infinispan-issues] [JBoss JIRA] (ISPN-9206) Handle long qualified region names more efficiently

[Sanne Grinovero (JIRA)]

    [ https://issues.jboss.org/browse/ISPN-9206?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13581762#comment-13581762 ] 

Sanne Grinovero commented on ISPN-9206:
---------------------------------------

With A+B and B+C you're talking about deployments and possibly how to distinguish (prefix?) deployment names. I guess that's maybe why we weren't understanding each other, I was talking about short names for the various entity names within a single Persistent Unit.

Those locally detected conflicts are the same thing I referred to in my first comment (I called them collisions though): I agree those are easy to fix for, with whatever strategy as long as it's deterministic so that other nodes would produce exactly the same alternative keys.

Yet if other nodes are expected by design to produce the same shortened keys, so when you connect to them... you are expecting them to have the same names. You can't distinguish this expected case from a conflict of keys - unless we premise it all with the assumption that applications need to be identical.

So it's easily proven that they need to be identical for any such optimizations to work - the alternative being that the cluster group runs a strong consensus protocol to establish short name mapping upfront, but that seems excessive IMO. 
We could just expect applications in the same cluster to be identical, and maybe have a simple consensus protocol which just verifies this.

> Handle long qualified region names more efficiently
> ---------------------------------------------------
>
>                 Key: ISPN-9206
>                 URL: https://issues.jboss.org/browse/ISPN-9206
>             Project: Infinispan
>          Issue Type: Enhancement
>          Components: Hibernate Cache
>    Affects Versions: 9.2.2.Final, 9.3.0.Beta1
>            Reporter: Radim Vansa
>            Assignee: Radim Vansa
>
> Hibernate region names are FQCNs, and when the prefix is added this can exceed the 255 byte ByteString limit. Also, it's inefficient to ship such long cache names with each command.



--
This message was sent by Atlassian JIRA
(v7.5.0#75005)