Re: 505 response a MUST?

> From: "Roy T. Fielding" <fielding@kiwi.ics.uci.edu>
> Date: Sun, 01 Mar 1998 22:30:52 -0800
> To: Scott Lawrence <lawrence@agranat.com>
> Cc: Larry Masinter <masinter@parc.xerox.com>,
>         HTTP Working Group <http-wg@cuckoo.hpl.hp.com>,
> jg@w3.org, http-wg@hplb.hpl.hp.com
> Subject: Re: 505 response a MUST?
> 
> >  That logic does not seem sound to me - if changing the major version
> >number means that the protocol _may_ be not backward compatible, then a
> >recipient that does not implement 2.0 cannot know whether or not it can
> >_correctly_ interpret the message as a 1.1 message.  Your logic assumes
> >that the authors of 2.0 will not make any semantic changes to existing
> >protocol elements.
> 
> No, it assumes a client won't send a 2.0 message to a 1.x server unless
> the 2.0 message is sufficiently semantically and syntactically compatible
> such that there is some reasonable assurance that the 1.x server will
> respond in an anticipated (and safe) manner, even if that means a 400
> response.  Forcing a server to respond in error just for the sake of
> an error only guarantees the worst-case behavior of an extra round-trip,
> and offers no extra safety in any case because deployed HTTP/1.x servers
> will never respond with 505.  That code is intended for future 2.x
> servers to say "piss-off" to older clients in the format of an HTTP/1.1
> message, since they will have to respond in the same major version as
> the old client.
> 
> ....Roy

I think Roy is right on this one; I'm removing the change from Rev-03.
				- Jim