- From: Scott Lawrence <lawrence@agranat.com>
- Date: Fri, 10 Sep 1999 08:46:07 -0400
- To: http-wg@hplb.hpl.hp.com
> The particular case I have in mind is an origin server > (which shall remain nameless) returning an entity body > with a 304. The client-side of our proxy is liberal > enough to cope with this on receipt, but the server- > side conservatively refuses to forward it. > > In this scenario it seems pretty reasonable to discard > the entity: Sounds right to me. > [...] But trying to generalize to other > kinds of upstream and downstream bad behaviour opens up > a can of worms. The options seem to be, > > 1. Forward broken HTTP. > > 2. Enforce strict conformance by responding to all bad > requests with 400s and all bad replies with 502s. > > 3. Fixup up bad requests/replies on their way through > wherever possible. > > (1) is pretty much out of the question; (2) is over the > top; and (3) is completely unspecified, so liable to be > done differently be different proxy implementors. I think that the only sensible thing to do is to figure each case out as they arise; you've arrived at a good solution for this one, and it sounds like you have it right - the server side is stricter than the client side. I would make sure that your log files include the Server: header for any offending responses, so that proxy operators can complain effectivly to vendors. I assume that you've not kept silent to the (nameless) vendor for this obvious bug. -- Scott Lawrence Director of R & D <lawrence@agranat.com> Agranat Systems, Inc. Embedded Web Technology http://www.agranat.com/
Received on Saturday, 11 September 1999 01:22:45 UTC