Re: PROPOSAL: i24 Requiring Allow in 405 Responses

Brian Smith wrote:
> Julian Reschke wrote:
>> here's what I'd like to do, based on Mark's summary and 
>> Stefan's latest
>> proposal:
>>
>> In the description of the Allow header, replace
>>
>>   This field cannot prevent a client from trying other methods.
>>   However, the indications given by the Allow header field value
>>   SHOULD be followed. The actual set of allowed methods is defined
>>   by the origin server at the time of each request.
>>
>> by
>>
>>   This field cannot prevent a client from trying other methods.
>>   The published set of allowed methods is defined by the 
>>   origin server at the time of each request. The absence of
>>   methods in this set has no defined semantics.
> 
> The first sentence "This field cannot prevent..." is meaningless and
> thus confusing. The server cannot prevent the client from doing
> anything. It should be removed.

Agreed.

> "The absense of of methods in this set has no defined semantics" implies
> that the presence of methods in the set does have (well-)defined
> semantics, which is not really true.

I wouldn't go that far; the presence does have semantics, with some caveats.

> How about:
> 
> "The methods listed in the Allow field of a response were allowed when
> that response was produced; additional, unlisted, methods may have been
> allowed at that time as well. Regardless of whether or not a method was
> listed in a previously-returned Allow field, the server MAY allow or
> refuse the use of any method in any (future) request. Servers are
> RECOMMENDED to include an accurate and complete set of methods in any
> Allow field returned in a response."

- This emphasizes a lot that the Allow set may change over time; in 
general I think this is misleading. It should only change when either 
the implementation or the "type" of the resource changes (for instance 
in WebDAV: collection vs non-collection)

- "RECOMMENDED" re-adds the normative requirement that most people want 
to get rid of, as some servers do not get this right in practice 
(remember that RECOMMENDED == SHOULD).

BR, Julian

Received on Tuesday, 18 March 2008 17:54:48 UTC