W3C home > Mailing lists > Public > www-html@w3.org > November 2006

Re: [OT] Suggestion: http request bundle

From: David Woolley <david@djwhome.demon.co.uk>
Date: Fri, 3 Nov 2006 22:39:25 +0000 (GMT)
Message-Id: <200611032239.kA3MdPk07671@djwhome.demon.co.uk>
To: www-html@w3.org

[ Off topic: IETF, not W3C ]
>   "Clients that use persistent connections SHOULD limit the number of
>    simultaneous connections that they maintain to a given server. A

This is just about good neighbourliness.  Violating it can hog resources
on the server and defeat the TCP congestion control algorithms.  Assuming
everyone uses the same conventions in this respect, there should be
no significant benefit over larger number of connections, as the larger
number of connections would run proportionately slower.

(Early implementations did start large numbers of connections.)

>    single-user client SHOULD NOT maintain more than 2 connections with
>    any server or proxy."
> which, implies that each subsequent request by the browser will be
> blocked until one of the two that are 'in progress' has completed.

As already pointed out, pipelining means that you can transmit all the
requests (or at least as many as will fit in the TCP window) before
receiving any of the responses, so the data flow rate is essentially
limited by the network bandwidth, not by round trip times for requests.

Therefore, the capability already exists, all you need to do is to 
make sure that servers, and, more of a problem, caching proxies support
it, and that browsers are configured to use it (IE through a proxy will
use HTTP/1.0, out of the box) and make effective use of it.
Received on Friday, 3 November 2006 22:51:07 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:06:14 UTC