[
https://issues.apache.org/jira/browse/PHOENIX-1501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14267200#comment-14267200
]
James Taylor commented on PHOENIX-1501:
---------------------------------------
I agree with Andrew's assessment of shims. Also, I agree this should be on a
case-by-case basis. Regarding the region co-location code (i.e.
IndexSplitTransaction), IMHO it'd be great if this lived in HBase. I know there
were some lingering doubts on this, but hopefully those can be solved, as
having the ability to co-locate regions is a very useful, general feature for
HBase.
> Remove or replace all uses of InterfaceAudience.Private HBase APIs and classes
> ------------------------------------------------------------------------------
>
> Key: PHOENIX-1501
> URL: https://issues.apache.org/jira/browse/PHOENIX-1501
> Project: Phoenix
> Issue Type: Task
> Reporter: Andrew Purtell
>
> Umbrella issue for removing or replacing all uses of
> InterfaceAudience.Private HBase APIs and classes.
> We won't be in a sustainable position if we don't establish a support
> contract with HBase for use of its private internals - each major HBase
> release will trigger potentially significant Phoenix refactoring.
> Let's approach it on a case by case basis. We can ask HBase to make private
> APIs and classes supported by promoting their audience to
> LimitedPrivate(PHOENIX). If that fails, we can negotiate a supportable
> interface and contribute or assist in the necessary refactoring. If that
> fails, we can redesign or implement internal analogues. If that fails, we can
> go back to the HBase community with a stronger argument for the necessity of
> a supportable interface and prevail.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
