- From: Klaus Weide <kweide@tezcat.com>
- Date: Mon, 11 Aug 1997 15:45:56 -0500 (CDT)
- To: Foteos Macrides <MACRIDES@sci.wfbr.edu>
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
On Mon, 11 Aug 1997, Foteos Macrides wrote: > Klaus Weide <kweide@tezcat.com> wrote: [ about the "OPTIONS probe" for non-compliant 1.0 proxies: ] > > > >It need not be a 200 status - any "HTTP/1.N" response with N>0 would > >be sufficient to detect that a next-hop server is not one of those proxies. > > > >My understanding is that no HTTP/1.1 server is required to implement > >any specific method. (So it could still be compliant if it responded > >to all methods with "HTTP/1.1 405 Method not allowed" or "HTTP/1.1 501 > >Not Implemented". Of course that's not very useful.) [ big snip ] > > As far as getting back 405 or 501 because that's possible based > on What is Written, I thought we were in LAST_CALL for moving HTTP/1.1 > from Proposed Standard to Draft Standard, hoping to catch any potentially > serious implementation problems, and interpretation of the Word is not my > forte. :) :) It would actually be reasonable behavior for a proxy-only server, i.e. a server which is not also configure to act as an origin server, to respond with an error status to all requests which don't use the absoluteURI form for Request-URI. And the non-absoluteURI form is required for the OPTIONS probe, to make sure it is really the next-hop proxy who answers. This leads to another question. Does the proposed OPTIONS probe actually *work* for an Apache/1.[23]* server acting as proxy? As Roy has revealed Apache-as-proxy is one of the cases where HTTP/1.1 response status is improperly forwarded. I suspect (but cannot test) that the same server would respond as HTTP/1.1 when checked with OPTIONS, making the probe worthless. Klaus
Received on Monday, 11 August 1997 13:47:56 UTC