Re: Comments on draft-ietf-http-negotiation-00.txt

Klaus Weide:
>
>On Wed, 19 Feb 1997, Koen Holtman wrote:
>
>> Klaus Weide:
>> >
>> >16/
>> >--- begin excerpt ---
>> >|13 Proxy support for transparent negotiation
>> >|
>> >
>> >Section 13 should include a part similar to 12.1 (maybe just a copy from
>> >there, with appropriate modifications).
>> 
>> I did copy some parts (and made a cut-and-paste error).  Which parts
>> do you think I should copy in addition?  I'm not aware of any
>> omissions.
>
>Most of the requirements may already be there scattered in other
>sections, but it would be nice to have them summarized here (parallel
>to 12.1).
>I am just taking a copy of 12.1 here, changing and annotating it,
>to give an idea what I meant:
>
>
>|13.1 Requirements for proxies [and caches?]
>
>   To implement transparent negotiation [a.k.a. `support t.c.n.',
>   a.k.a. `be compliant with this specification'?] on a resource, a proxy
>   must be able to [?????] when getting a GET
>   request on the resource.  It should also be able to [?????]
>   for HEAD requests.  A list response must
>   always be sent [or the request be forwarded to the origin server?]
>   if the request includes a Negotiate header with only
>   a "trans" directive.
[.....]

OK, I think I see what you mean now.  

Requirements like the ones above are really implied by the general rule that
a proxy may not do antyhing unless this is explicitely allowed by a spec.  A
HTTP proxy cache is basically a transparent HTTP relay onto which some
optimisations _may_ be added, if these optimisations are specifically
allowed for by the spec the proxy adheres to.

Taking your last sentence above:

> A list response must
>   always be sent [or the request be forwarded to the origin server?]
>   if the request includes a Negotiate header with only
>   a "trans" directive.

This requirement will be satisfied automatically if the proxy plays by the
rules.  The proxy does not even have to know what a list response is if it
does not wish to implement the specific optimisations for it in the TCN
draft.  So a proxy need not contain specific code to satisfy this
requirement.  It would therefore be confusing to proxy implementers if we
would list the requirement explicitely.

I'll add an explanation of these issues to the next draft.  Please check if
the text in the next draft resolves your concerns.

>  Klaus

Koen.

Received on Monday, 24 February 1997 11:43:14 UTC