- From: Danny Ayers <danny.ayers@gmail.com>
- Date: Fri, 5 Oct 2007 18:00:33 +0100
- To: "Linking Open Data" <linking-open-data@simile.mit.edu>
- Cc: "SW-forum Web" <semantic-web@w3.org>
On 05/10/2007, T.Heath <T.Heath@open.ac.uk> wrote: > > BTW, this isn't meant as a nay-saying bit, but as a wanting > > to derive useful lessons from past experience. The above links *do* > > contain nay-saying, but I'm less interested in refuting that than > > understanding what drove it and if there is any way to do better. I blame XML! Not entirely joking - there's a great quote, the attribution of which I keep losing, to the effect that "what's new about the Semantic Web isn't the 'Semantic' but the 'Web'". Ease back a few notches and what was new about HTML wasn't that it was a simple document markup format, but that it supported global linking. But XML's success has really come from the doc side. Documents are clearly pretty significant when it comes to the Web, and XML provides a means to expressing data in document form relatively easily. But there's no inherent connection between those two statements (though namespaces make a big difference). I reckon it's significant that a lot of other RDF crit (including some of this XUL material) boils down to the fact that for a lot of applications, graph-shaped data isn't needed and trees or lists are more than adequate. Even in HTML it's the tree structure that's in your face if anything, not the graph. Generalised graph serializations are never going to be pretty, and add to this the need to URI-qualify names and things start looking worse. Kingsley mentioned RDF/XML - yup. Folks like Bill deHora have been saying forever that RDF/XML is RDF's Achilles Heel, and I reckon they're right up to a point. Given the choice between a universal representation that's ugly & confusing and a single-purpose data representation that fits the data and is easy to comprehend (because we wrote it ourselves...), a typical developer will likely opt for the latter. It seems like it's too late to save RSS and Moz from their non-RDF paths, but in each case I suspect the road ahead leads back to RDF-style linkiness. RSS gets cleaned up to Atom, Atom gets passed around using Web-friendly AtomPub, the link prevails. Moz gets SQLite inside, but given that a browser is a tool for the Web, the link is bound to prevail. Coming from a data perspective, the value of linked data on the Web is non-obvious, but I reckon once someone notices it, there's not turning back. From a doc perspective, microformats solve some of the immediate problems for certain classes of data, but more generally RDF (or something very similar) is needed. Now I'm rambling. I guess what I'm trying to say is that some of the earlier RDF efforts tried to do too much too soon, and while the Big Picture solution might have been pretty much on the nail, the implementation details for the necessary components left a lot to be desired. There's also the injection "from above" side Bijan mentioned - I'm not sure how widespread this has been, but the perception seems pretty common for the pushback when RDF is suggested to be as if it's being imposed. Certainly some of the traditional problems which can be solved technically, have been solved technically - Turtle syntax is legible, tools are making life easier, and GRDDL significantly reduces the up-front commitment needed to adopt RDF (pretty handy socially & politically too). Purpose-specific XML can be joined to the Web post hoc. No doubt a lot of mistakes have been made with evangelism (I'm as much to blame as anyone) - as I think LeeF and/or EliasT were stressing a year or so back, showing is much more effective than telling. If the intuitions that a Web of Data is a good idea and the RDF approach is a good one are right, then an indicator would be a significant increase in developers coming to RDF "of their own accord", without the need for outreach. But then again there's always the possibility of reinvention, so I guess we can't just sit back and watch the show... Cheers, Danny. -- http://dannyayers.com
Received on Friday, 5 October 2007 17:00:43 UTC