W3C home > Mailing lists > Public > ietf-http-wg@w3.org > April to June 2008

Re: PROPOSAL: i99 Pipelining Problems

From: Henrik Nordstrom <henrik@henriknordstrom.net>
Date: Mon, 07 Apr 2008 10:19:58 +0200
To: Mark Nottingham <mnot@mnot.net>
Cc: "Roy T. Fielding" <fielding@gbiv.com>, Julian Reschke <julian.reschke@gmx.de>, HTTP Working Group <ietf-http-wg@w3.org>
Message-Id: <1207556398.21800.173.camel@HenrikLaptop>

mån 2008-04-07 klockan 12:20 +1000 skrev Mark Nottingham:
> "Some early implementations of HTTP/1.1 servers and proxies have been  
> noted to implement pipelining incorrectly, and some commonly-deployed  
> (if not spec-compliant) devices may interfere with its correct  
> operation. Clients choosing to send pipelined requests on the open  
> Internet should, as a result, do so cautiously."

I prefer not adding this to the specs as it reduces the push on such
vendors to get their software fixed to support pipelining, opening up a
for dialogue like the following

Client vendor: Your server appears to be broken and fails when we send
pipelined requests to it.

Server vendor: Your application is the only client application we know
of having this problem, and specs says you should be cautious when
sending pipelined requests so you deal with it. 

Yes, this shows how much trust I have in the industry on getting things

Comments aiding implementers in how to deal with the current mess of
clearly broken and non-compliant implementations is best placed in a
separate informal document outside the standard documenting known bugs
and how to deal with them. The focus in the standard text itself should
be longterm interoperability. These implementations is very likely to
decline over time, and having such comments in the standard text itself
only adds confusion.

So far there is no evidence that the amount of broken implementations is
so high that implementing pipelining isn't possible or useful. In fact I
would say the opposite has already been proved in the wild with several
non-browser applications making very successful use of pipelining. It's
true that enabling pipelinng is hard for the major browser vendors as
their users expect them to deal with pretty much every crappy server out
there no matter how broken it is, but thats a position they have placed
themselves in.  I would be very glad the day the major browsers started
to actually alert the user when a broken server is detected instead of
just silently work around it, placing some pressure on getting servers

Received on Monday, 7 April 2008 08:23:01 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 11:10:45 UTC