[
https://issues.apache.org/jira/browse/COUCHDB-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12976878#action_12976878
]
Filipe Manana commented on COUCHDB-1008:
----------------------------------------
Randall,
I don't think we need that nodelay option deprecation message. As far as I can
see, the nodelay option under the section [httpd] was added for 1.1.0, which
means it's not used in any official release. Couldn't find any reference to it
in 1.0.x and 0.11.x
About the nodelay option, it's confusing that it's both in server_options and
socket_options. I think I would prefer to leave it in socket_options. The
mochiweb nodelay option is applied to the listen socket. When an 'accept' is
done, the returned socket inherits the listen socket's configuration . Since
the socket options are applied at the beginning of the loop, if nodelay is
specified in the socket_options, it will override the nodelay setting specified
in server_options.
Maybe others have a different point of view.
thanks
> Add socket_options to the ini .configuration
> --------------------------------------------
>
> Key: COUCHDB-1008
> URL: https://issues.apache.org/jira/browse/COUCHDB-1008
> Project: CouchDB
> Issue Type: Improvement
> Reporter: Filipe Manana
> Assignee: Filipe Manana
> Attachments: COUCHDB-1008.patch, COUCHDB-1008.patch
>
>
> Under some scenarios it's useful to set custom socket options, namely the
> recbuf and sndbuf options.
> For example, setting the recbuf option to 256Kb, dropped the time to upload
> a 100Mb attachment from 2507499us to 1387625us. The default value of this
> parameter is 8192 bytes (set by mochiweb).
> The following article explains how some socket options might boost
> performance under some common scenarios:
> http://www.ibm.com/developerworks/linux/library/l-hisock.html
> Patch attached
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
