[
https://issues.apache.org/jira/browse/HADOOP-7379?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13048451#comment-13048451
]
Todd Lipcon commented on HADOOP-7379:
-------------------------------------
bq. What will it take to get ipc.Server to understand PB rather than Writable?
You mean using PBs for the actual invocations? Or the "header" type info at the
beginning? I think most of the value from PB comes from the evolvable parameter
types, but moving to PB for the headers would also allow us to do things like
add extra metadata to requests without breaking previous versions (eg
dapper-style tracing info?)
> Add ability to include Protobufs in ObjectWritable
> --------------------------------------------------
>
> Key: HADOOP-7379
> URL: https://issues.apache.org/jira/browse/HADOOP-7379
> Project: Hadoop Common
> Issue Type: Improvement
> Components: io, ipc
> Affects Versions: 0.23.0
> Reporter: Todd Lipcon
> Assignee: Todd Lipcon
> Fix For: 0.23.0
>
> Attachments: hadoop-7379.txt, hadoop-7379.txt
>
>
> Per HDFS-2060, it would make it easier to piecemeal switch to protocol buffer
> based data structures in the wire protocol if we could intermix the two. The
> IPC framework currently provides the concept of "engines" for RPC, but that
> doesn't easily allow mixed types within the same framework for ease of
> transition.
> I'd like to add the cases to ObjectWritable to be handle subclasses of
> {{Message}}, the superclass of codegenned protobufs.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
