[
https://issues.apache.org/jira/browse/HBASE-13902?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14586231#comment-14586231
]
stack commented on HBASE-13902:
-------------------------------
After a quick skim...
Radical!
Its easy to review (smile) given its just removal (and a one line change)
How can we get away with removing such fundamental types as Call and
RpcClientImpl Want to say a few words on it [~jurmous] ?
The plus is now we have an async chassis and so can do an async Table -- that
is great -- and do we get a simplification of the client path but can you think
of any downsides? For example, doing sync calls on an async chassis continue
to be slower than pure sync (I'd guess so but probably not by much?). I could
help test this out if that'd help. Any other downsides you can think of ?
> Remove Sync RpcClientImpl
> -------------------------
>
> Key: HBASE-13902
> URL: https://issues.apache.org/jira/browse/HBASE-13902
> Project: HBase
> Issue Type: Improvement
> Components: Client
> Reporter: Jurriaan Mous
> Assignee: Jurriaan Mous
> Attachments: HBASE-13902.patch
>
>
> For an async Table api to be supported we need to remove the sync
> RpcClientImpl.
> For the Async Table Api some internals for scanning are changed to use async
> apis so they can't be used with the Sync RpcClientImpl.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
