Re: [ietf-types] Request for review application/gml+xml

[Julian Reschke <[e-mail]>]

On 2012-01-11 17:50, Jan Algermissen wrote:
>
> On Jan 11, 2012, at 10:46 AM, Julian Reschke wrote:
>
>> On 2012-01-11 10:22, Jan Algermissen wrote:
>>>
>>> On Jan 11, 2012, at 9:38 AM, Julian Reschke wrote:
>>>
>>>> On 2012-01-11 08:05, Jan Algermissen wrote:
>>>>> Carl,
>>>>>
>>>>> On Jan 11, 2012, at 2:54 AM, Carl Reed wrote:
>>>>>
>>>>>> "version": If provided, this parameter indicates the GML version
>>>>>>     used in the GML document. Only the major and the first minor
>>>>>>     version number are provided, e.g. "3.2".
>>>>>>
>>>>>>     The value may be provided with or without quotes. i.e.,
>>>>>>     application/gml+xml; version=3.1 shall be treated as identical to
>>>>>>     application/gml+xml; version="3.1".
>>>>>>
>>>>>>     In cases where elements from multiple GML versions are used,
>>>>>>     the parameter shall indicate the highest GML version used in the
>>>>>>     document.
>>>>>>
>>>>>>     The parameter can be used to provide protocol-specific operations,
>>>>>>     such as version-based content negotiation in HTTP
>>>>>
>>>>> I cannot find the reference just now, but if I am not seriously wrong, media type parameters are not taken into account by HTTP conneg. (How could they, given that their meaning is defined by the media type itself except for q and charset et al.
>>>>> ...
>>>>
>>>> Why would that be a problem?
>>>
>>> I think that the correct way of implementing conneg is to strip all media type specific parameters before matching (otherwise the code would need to be aware of the media type specific parameters).
>>
>> But that's not what is specified in HTTP.
>
> Can you point me to that part of the spec? I saw the 'format' example in your reference to media type ranges, but I cannot (personally) read that the way you seem to do.
>
>>
>>> The proposed text above suggests that the version parameter would be recognized by such implementations - which it would not. I.e.
>>>
>>> Accept: application/gml+xml; version=3.1
>>>
>>> would maybe yield
>>>
>>> Content-Type: application/gml+xml; version=1.2
>>
>> I think that would be a bug according to the HTTP spec.
>
> Doh, I am shocked - if you are right then I have lived with a very clouded spot of HTTP in my brain for years.
>
> So you are saying, if there are parameters (besides q) that do not *exactly* match conneg should say 'Not Acceptable'?

Well, it's all advisory; so there is wiggle room.

> IOW, this would also not be correct:
>
> Accept: application/gml+xml; version=3.1
>
> Content-Type: application/gml+xml; version=3.5
>
> And also this would be incorrect:
>
> Accept: application/atom+xml;type=entry
>
> Content-Type: application/atom+xml
>
> Or am I misunderstanding you?

No, that's my understanding. But of course I may be wrong; I assume this 
is a feature that hasn't been used a lot yet.

Best regards, Julian