Re: #147: header-specific canonicalisation

On Mar 4, 2010, at 12:49 AM, Julian Reschke wrote:

> On 04.03.2010 09:17, Roy T. Fielding wrote:
>> On Mar 3, 2010, at 5:05 AM, Julian Reschke wrote:
>>> Note:
>>> 
>>>   If a header field is absent from a request, it can only match another
>>>   request if it is also absent there.
>>> 
>>> Is that still true? I don't think; maybe we can remove the sentence completely?
>> 
>> Yes, it is still true.
>> 
>>    Accept:
>> 
>> means accept nothing (give me a 300 instead).  No Accept header means
>> accept everything.
> 
> I do agree for "Accept".
> 
> But a header *could* have a normalization behaviour, where the presence of an empty header and the absence of the header field have the same semantics. (Not a good idea, probably, but that doesn't matter here).
> 
> In that case, a header-field specific normalization should allow to treat them as matching, no?

No, there are no request header fields with that characteristic
that would appear in Vary, and we are better off letting caches
be ignorant of the meaning of a field while running these tests.

....Roy

Received on Thursday, 4 March 2010 20:13:41 UTC