[
https://issues.apache.org/jira/browse/HADOOP-4853?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12655958#action_12655958
]
Steve Loughran commented on HADOOP-4853:
----------------------------------------
+1
Regarding the round trip, isn't there that feature of TCP whereby the server
accepting the connection can include some data in the response; the first
packet sent back after the connection is set up doesn't have to be empty. The
server could include some version info and its part of the authentication
process (a list of approved auth mechanisms and a seed value?).
> Improvement to IPC
> ------------------
>
> Key: HADOOP-4853
> URL: https://issues.apache.org/jira/browse/HADOOP-4853
> Project: Hadoop Core
> Issue Type: Improvement
> Components: ipc
> Affects Versions: 0.20.0
> Reporter: Arun C Murthy
> Assignee: Arun C Murthy
>
> I'd like to propose an improvement for consideration given my experience of
> working on HADOOP-4348:
> Currently the first call doubles up as a 'connection setup' trigger. I'd like
> to propose adding a new 'init' call which is always called by the clients for
> connection setup. The advantages are:
> * We could fold in the getProtocolVersion call into the setup call, this
> ensures that the Server always checks for protocol versions, regardless of
> whether the (malicious?) client does an explicit call for getProtocolVersion
> or not.
> * We could authorize the connection here.
> * We could add to check to ensure that the Server instance actually
> implements the protocol used by the client to communicate, rather than fail
> on the first IPC call
> The flip side being an extra round-trip.
> Lets discuss.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
