Re: comments on draft-nottingham-http-link-header-04

Ah, you've rumbled my secret plan for a link relation caching startup;  
ah well, there's still DTDs, XML Namespaces and the like out there.

I'll have a think about this; I agree that SHOULD is too strong, but  
it would be nice to allow the semweb people to do what they need to if  
they really want to. At the moment, I'm thinking about making the  
entire thing non-normative; i.e., shoulds instead of SHOULD. I'll try  
to float some proposed text in a little while.s


On 27/02/2009, at 1:45 AM, Julian Reschke wrote:
>
> Section 4.2:
>
> "An extension relation type is a URI [RFC3986] that, when  
> dereferenced, SHOULD yield a document describing that relation type."
>
> It seems to me that "SHOULD" is way too strong here:
>
> 1) As that document is not required in any way to process the link  
> relation, not providing it won't affect interoperability at all.
>
> 2) It also seems to invite dereferencing, which I expect many will  
> oppose.
>
> Proposal: rephrase this in a way similar to RFC 3648, Section 5.1:
>
> "For collections that are ordered, the client SHOULD identify the  
> semantics of the ordering with a URI in the Ordering-Type  
> header, ... Setting the value to a URI that identifies the ordering  
> semantics provides the information a human user or software package  
> needs to insert new collection members into the ordering  
> intelligently. Although the URI in the Ordering-Type header MAY  
> point to a resource that contains a definition of the semantics of  
> the ordering, clients SHOULD NOT access that resource to avoid  
> overburdening its server. ..." -- <http://greenbytes.de/tech/webdav/rfc3648.html#rfc.section.5.1 
> >


--
Mark Nottingham     http://www.mnot.net/

Received on Thursday, 26 February 2009 22:18:25 UTC