- From: Julian Reschke <julian.reschke@gmx.de>
- Date: Tue, 12 Feb 2008 19:01:44 +0100
- To: Geoffrey Sneddon <foolistbar@googlemail.com>
- CC: ietf-http-wg@w3.org, Mark Nottingham <mnot@mnot.net>
Geoffrey Sneddon wrote: >> The reason we are making a change is that user agents indeed do *not* >> do what RFC2616 said. Thus we're removing that specific requirement. >> IMHO. > > My point is that the issue with a MUST level requirement is that because > of what other specifications require (in the text/xml case, RFC3023) it > is still impossible to satisfy HTTP's requirements, therefore user > agents still do *not* (and will not) do what HTTP says. The four most I think RFC3023 is making the requirement *because* of the default in RFC2616. So it was RFC2616 overruling RFC2046, in turn being overruled by RFC3023. > popular browsers aren't going to change their behaviour regarding this. > The only sensible thing to do, for the sake of interoperability, is to > match what they do, which we still don't do. As Roy just stated, there are many other clients we need to consider in addition to the "four most popular browsers". What may seem to be the best behavior for them may be unacceptable to other user agents (and this includes many things that are not browsers at all). What we need to discuss now is whether we want HTTP/1.1 to continue overruling RFC2046, whether we want to start overruling RFC3023 as well, and so on. My preference clearly is to get out of that business, and let the MIME specs rule, in particular as those aforementioned browsers did not do what HTTP/1.1 says anyway. So is there anything we *can* specify that matches reality? BR, Julian
Received on Tuesday, 12 February 2008 18:02:10 UTC