[
https://issues.apache.org/jira/browse/THRIFT-23?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12600219#action_12600219
]
David Reiss commented on THRIFT-23:
-----------------------------------
Sorry for closing this. I wasn't paying close attention. My comment on the
Ruby version applies to any language, though:
I'm definitely fine with including this, but I should warn you in advance that
one of my coworkers did some experiments and was unable to find any significant
performance difference between PF_UNIX and loopback TCP. Our conjecture was
that the TCP stack has gotten a lot of optimization, while the less-used
Unix-domain stack is more primitive. However, this test was only on one version
(2.6.12, I believe) of one OS (Linux), so YMMV.
> TUnixSocket
> -----------
>
> Key: THRIFT-23
> URL: https://issues.apache.org/jira/browse/THRIFT-23
> Project: Thrift
> Issue Type: New Feature
> Components: Library (Java)
> Reporter: Bryan Duxbury
> Priority: Minor
>
> At Rapleaf, we'd love to have a TUnixSocket and TUnixServerSocket that ran in
> Java. Before someone chimes in and says, "Java doesn't support Unix
> sockets!", I would point to JRuby's RubyUnixSocket/Server classes. They use
> JNA to get down to the bare metal and make it work. If we could take
> advantage of that class, then we'd have a really fast local machine
> transport.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
