[jira] [Commented] (PHOENIX-1593) Use protobuf for all client/server data interchange in next major release

["James Taylor (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/PHOENIX-1593?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14303976#comment-14303976 ] 

James Taylor commented on PHOENIX-1593:
---------------------------------------

Thanks, [~anoop.hbase]. Take your pick :-) Some of these might end up sharing the same protobuf message depending on how we want to decompose them. We have some "hackery" that piggybacks on our UngroupedAggregateRegionObserver (when you do a SELECT count(*) FROM T) and sets various "special" attributes to do various things when the scan occurs. It's possible that some of these could be folded into the same protobuf. Some up front design here would be beneficial.

> Use protobuf for all client/server data interchange in next major release
> -------------------------------------------------------------------------
>
>                 Key: PHOENIX-1593
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-1593
>             Project: Phoenix
>          Issue Type: Bug
>            Reporter: James Taylor
>
> Some things are already protobuffed, like PTable. However, other things aren't, such as Expression serialization along with various other serialization we do through Scan/Put attributes for Phoenix operations.
> In the next major release, 5.0, we should firm this up and have a single protobuf message per operation (it's too late to do it for the 4.x branch, as we'd by definition be breaking b/w compatibility). This will prevent us from having to hack around the way we do to prevent client/server backward compat issues.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)