- From: Amos Jeffries <squid3@treenet.co.nz>
- Date: Mon, 25 Jul 2011 15:12:35 +1200
- To: ietf-http-wg@w3.org
On 25/07/11 13:29, Adrien de Croy wrote: > > > On 25/07/2011 6:31 a.m., Mark Nottingham wrote: >> On 24/07/2011, at 2:11 PM, Willy Tarreau wrote: >> >>> On Sun, Jul 24, 2011 at 02:06:17PM -0400, Mark Nottingham wrote: >>>> <http://trac.tools.ietf.org/wg/httpbis/trac/ticket/78> >>>> >>>> Proposal: >>>> >>>> 1) Clarify that WWW-Authenticate can appear on any response, and >>>> that when it appears on any other than a 401, it means that the >>>> client can optionally present the request again with a credential. >>> Does this mean it's only for other 4xx or for any status ? It might have >>> implications with non-idempotent requests if a client can repost a >>> request >>> that led to a 200 for instance. >> Any status. Good point about non-idempotent requests; we'll need to >> make clear it's not about automatically retrying requests, but instead >> that sending the same request with credentials might have a different >> affect. > > isn't this redundant? > > I see requests with credentials all the time, when no previous > WWW-Authorize had been sent in any response. I disagree on the redundant. This clarifies and documents the existing behaviour which must be handled by implementers to improve compatibility. Not all of us see these same requests you do. > > So clients are already taking any liberties they like to send > credentials when they please. I don't know that it adds anything to HTTP > to explicitly tell them they may do this in protocol. They are doing it > anyway. > > Otherwise are we going to prohibit the sending of creds when no > WWW-Authorize had been sent? Prohibiting that would be a bad idea. It is possible for the client agent to use non-HTTP means to verify the connection security requirements before sending credentials over it. Likewise the mere presence of a WWW-Authorize challenge means nothing about the security of the channel. Despite NTLM assumptions to the contrary. A mention in the security considerations should be enough there. AYJ
Received on Monday, 25 July 2011 03:13:11 UTC