Re: Link header is representation metadata?

But if you look at RFC 5988, most of the relations seem to be
*resource* relations, not representation relations. The introduction
speaks of "relationships between resources"; also

   "links between resources
   need not be format specific; it can be useful to have typed links
   that are independent of their serialisation, especially when a
   resource has representations in multiple formats."

So it's a bit confusing to say that Link: is an entity-header (which,
I admit, 5988 does). I guess a "next" Link: on a 404 response would
tell you what the next resource is after one that the representation
carried in the 404 response is a representation of, not the next one
after the requested resource (which may still exist somewhere).

This rules out one application I really wanted to use Link: for, which
is to use it with a 301 or 302 response to give information about the
resource identified by the URI. You could call this a "value-added
redirect". With this interpretation it would refer to some resource of
which the entity in the redirect response was a representation - a
very different beast.  Oh well!

Jonathan

On Tue, Feb 1, 2011 at 4:24 PM, Nathan <nathan@webr3.org> wrote:
> Roy T. Fielding wrote:
>>
>> On Feb 1, 2011, at 11:50 AM, Nathan wrote:
>>>
>>> Julian Reschke wrote:
>>>>
>>>> On 12.10.2010 01:02, Nathan wrote:
>>>>>
>>>>> I'm looking for a quick bit of guidance on whether the Link header
>>>>> would
>>>>> class as representation metadata
>>>>
>>>> The Link header spec says "entity header", which is the terminology we
>>>> used in HTTPbis until draft 10; so I think it's supposed to fall into the
>>>> same class as "Content-Language", for example.
>>>
>>> Sorry but I need to come back to this one, can I get a definitive answer
>>> (please :)) on whether the Link header is representation metadata (like
>>> Content-Type), or not?
>>
>> It is sometimes metadata. Some of those times it is representation
>> metadata, like Content-Type, whereas in other times it can be
>> resource metadata, like Vary.
>>
>> And occasionally it is just part of the representation, though that
>> tends to confuse people who think data == body and headers == meta.
>>
>> What the link is for depends entirely on the rel value.
>
> Thanks Roy, that's exactly the answer I was hoping for!
>
> Best,
>
> Nathan
>
>

Received on Sunday, 6 March 2011 15:38:11 UTC