W3C home > Mailing lists > Public > public-webapi@w3.org > February 2008

Re: Pipelining Control Proposal

From: Kris Zyp <kzyp@sitepen.com>
Date: Tue, 19 Feb 2008 15:50:11 -0700
Message-ID: <0cc401c87349$c94b6ef0$4200a8c0@kris>
To: "Anne van Kesteren" <annevk@opera.com>
Cc: "Web API WG \(public\)" <public-webapi@w3.org>

> I think this out of scope for XMLHttpRequest. (In fact, earlier versions 
> of the XMLHttpRequest suggested that implementations, if pipelining is 
> supported, should use pipelined connections.) Pipelining is something that 
> all network traffic can benefit from, not just XMLHttpRequest.

Of course, most network traffic can benefit from pipelining... However XHR 
may well become the exception if the target connection can't be controlled. 
It might not seem like it right now, but with the increasing usage of 
long-lived responses, not being able to control pipelining could result in 
requests that become queued behind responses that are never fulfilled 
because they are forever streaming. Pipelining could become an increasing 
hazard in the next generation of webapps. On the otherhand if we have 
pipelining control, it could become an incredibly valuable tool to not only 
prevent pipelining on dead-end connections, but also build efficient duplex 
channels of communication on those same connections. Also, it is important 
to note that this is not so much about ordering pipelining to happen or not 
to happen, but about hinting at which connection should be used when 
pipelining does happen. This is hardly a large implementation burden, with 
very large benefits down the road.
Kris
Received on Tuesday, 19 February 2008 22:51:22 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 19 February 2008 22:51:24 GMT