- From: Nathan <nathan@webr3.org>
- Date: Mon, 05 Apr 2010 00:32:55 +0100
- To: Jan Algermissen <algermissen1971@mac.com>
- CC: www-tag@w3.org
Jan Algermissen wrote: > On Apr 5, 2010, at 1:17 AM, Nathan wrote: > >> Hi All, >> >> I've hit upon something which may be a future issue (unsure). >> >> As the read/write web of data is realised (and assuming that RDF remains >> the primary datatype for linked data), then machines will become reliant >> on specific ontologies in some use-cases. >> >> With XML we have things like Atom - application/atom+xml. However if >> Atom were in RDF instead, and any serialization could be used >> (n3/rdf/xml etc), then how would one create a media type for it? >> >> Major / commonly used ontologies will arise; just as with the many >> registered ****+xml media types, there may be a need for ****+rdf but >> without the limitation of a specific serialization, or with the addition >> of multiple serializations. >> >> Examples: Machines / Agents may wish to indicate they "Accept:" a >> specific ontology "i understand x ontology / type of data in y&z >> serializations" - perhaps foaf-onto+rdf on the client side or >> diff-onto+rdf on the server side (accept-patch). >> >> Perhaps a non-issue, but worth mentioning I hope, > > I usually address the subclassing issue (which is primarily an issue > of expressing a client side capability of expectation, IMHO) with > a profile parameter on the media type(s) in the Accept header: > > Accept: application/rdf+xml;profile=http://xmlns.com/foaf/0.1/ > > The server can respond with just Content-Type: application/rdf+xml or, > if desired, with a 406 Not Acceptable if the profile 'request' cannot > be satisfied. > Fantastic that about answers my question :) will need to check how that stacks up w/ Accept-Patch from the new HTTP PATCH rfc; if so then I'm all good (personally). The only thing I will add, that later realised the above is primarily due to the multiple serializations of rdf; if only a single rdf serialization existed then it'd be a non issue. As even with the above profile=, We'll still need to chain up 2-3 of them for each serialization / preference. Regards, and thanks - that's enough to get me going! Nathan
Received on Sunday, 4 April 2010 23:33:37 UTC