HTTP Pipelining or batchhttp

1 view
Skip to first unread message

Jeremy Dunck

unread,
Sep 18, 2009, 3:21:46 PM9/18/09
to fluiddb...@googlegroups.com
It seems to me that a common app use case for fluiddb will consist of
something like:

1) "select all objects which have this tag"
2) followed by local processing logic, followed by
3) "update these many other objects with these other tags".

The third step takes lots of separate requests to fluiddb.

Have you considered pipelining?
http://code.activestate.com/recipes/576673/

Or batchhttp?

http://sixapart.github.com/batchhttp/
http://martin.atkins.me.uk/specs/batchhttp

Terry Jones

unread,
Sep 20, 2009, 8:04:56 PM9/20/09
to fluiddb...@googlegroups.com
Hi Jeremy

>>>>> "Jeremy" == Jeremy Dunck <jdu...@gmail.com> writes:
Jeremy> 1) "select all objects which have this tag" 2) followed by local
Jeremy> processing logic, followed by 3) "update these many other objects
Jeremy> with these other tags".

Jeremy> The third step takes lots of separate requests to fluiddb.
Jeremy> Have you considered pipelining?
Jeremy> Or batchhttp?

We've thought about doing something like this, but I wasn't aware of either
of pipelining or batchhttp. If we were going to support one, I think
batchhttp would probably be the simplest.

Thanks a lot. I don't know when we'll get there, but it's nice to know
there are standards emerging for this.

Terry

Reply all
Reply to author
Forward
0 new messages