- From: Danny Ayers <danny666@virgilio.it>
- Date: Fri, 21 Nov 2003 10:40:31 +0100
- To: "Jim Hendler" <hendler@cs.umd.edu>, "Joshua Allen" <joshuaa@microsoft.com>, "Patrick Stickler (NMP-MSW/Tampere)" <patrick.stickler@nokia.com>
- Cc: "Graham Klyne" <GK@ninebynine.org>, "Dan Brickley" <danbri@w3.org>, <www-rdf-rules@w3.org>, <www-rdf-interest@w3.org>
> >> But despite the fact that we > >> > don't even have the slightest freekin' semblance of a consistent > >> > data-access architecture, > >> > >> We have the web. > > > >PUT/GET are a data-access architecture for documents, period. RDF is > >not a document data model; it's a graph data model. If you want to shim > >a document data model on top of your graph storage, then fine -- but you > >better figure out how to interact with your graph storage first. Until > >you have a consistent way of accessing your graph data model, you're > >going to have a post-babel chaos. > > > Sorry, just trying to change the name on this thread now that it no > longer is really related to the issue of whether to form WGs at this > time and if so which ones... Fair enough. Re. your original post, I agree entirely with your proposal, especially: "1 - I think there is a clear and present need in the RDF community for a way to essentially request a set of triples from a remote store -- essentially an RDF remote access API." Where PUT/GET (and POST, DELETE) figure in this are as the existing web's network API. I would suggest that a simple application of this to the RDF model would yield the required remote access API. Patrick and I have been arguing over implementation, but I believe that the functionality offered by URIQA [1] is just the kind of thing we should be aiming for in the first instance. There is a need for a standardization of sophisticated, fine-grained access to graph models, but I think as Jim was proposing, this isn't the immediate requirement. On Joshua's point, PUT/GET etc are a data access architecture for representation of resources on the web. True, it's currently used primarily with documents, but if the system is intended to use the web then it makes sense to leverage what we've already got. The basic difference between Joshua's and my position here is that I don't think the querying of RDF models on the web should be orthogonal to the web architecture. The reason I picked up on this thread is that I believe there is a danger that an approach might be taken that doesn't mesh well with the existing web: SQL-like queries could be posted around in SOAP packets, but this would be largely opaque to the rest (or REST) of the web. Patrick's approach is in a different direction, one which certainly does take into account the existing web, I'm just not convinced there isn't something closer. Cheers, Danny. [1] http://sw.nokia.com/uriqa/URIQA.html
Received on Friday, 21 November 2003 04:48:45 UTC