RE: repeat of a comment from a long time ago

I think it should say MUST, actually. Since we don't know anything about
what the new methods do, we don't know that it is safe to pass them
along. For example, they may require that something be done to the
contents of the cache in order to interact properly with other new
methods, or even with existing ones.

I can think of one way to make it safe(r) -- to go into tunnel mode, for
both the request with the new method, and all subsequent requests on
that resource. That's not too practical, and doesn't help if the new
method affects collections of related resources.

I would agree with user-extensibility only if there is some way, that we
could spec in HTTP 1.2, for the method to declare that it's OK for
proxies to just forward the request toward the origin server.

Paul
>----------
>From: 	Shel Kaphan[SMTP:sjk@amazon.com]
>Sent: 	Monday, June 03, 1996 11:11 AM
>To: 	http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
>Subject: 	repeat of a comment from a long time ago
>
>
>In 5.1.1, it says:  "Servers SHOULD return the status code ... 501
>(Not Implemented) if the method is unrecognized or not implemented by
>the server."
>
>At least this isn't "MUST", but I don't even see the SHOULD as valid
>for proxies passing methods they don't recognize.  I'm not suggesting
>that proxies be required to forward requests they don't recognize,
>merely that they be _allowed_ to without violating any recommendations
>of the spec.   I haven't gotten the sense that there is a consensus
>opinion
>about how user-extensible HTTP should be on request-method, but it
>would be good
>to leave it open until there is such a consensus.
>
>This could be fixed simply by using the words "origin servers" instead
>of
>"servers" in this section.
>
>Shel Kaphan
>sjk@amazon.com
>
>

Received on Monday, 3 June 1996 11:52:36 UTC