Re: Is promoting RDF+XML a lost cause?

Yes to everything ;-)

Certainly for people coming to RDF from XML with preconceptions of how
to solve problems from that space, RDF/XML is a big problem. The cost
of the syntax looms large long before the developer sees the potential
benefits of the model. But RDF/XML does work pretty nicely for
machine-machine comms, is in the specs and isn't going to go away any
time soon.

re. XML tool compatibility with RDF/XML, yep it is a problem, but
don't overestimate tool standardization within XML. Both SAX and DOM
seem to be sidestepped by many developers these days, with pull
parsers and tree-based APIs that exploit individual programming
language idioms (Did I see a gay & lesbian version of DOM the other
day? Maybe I dreamt it.)  There's something of a mismatch between the
goals of XML Schema and RDF serializations, but many XML folks seem to
be tending towards alternatives like Relax NG anyhow. XSLT is a big
sticking point, but even then I suspect there's a danger of
overstating the problems. Given that James & Ian's have a C# RDF/XML
parser [1] actually based around XSLT there may well be avenues
forward that haven't yet been explored. I personally really like XSLT,
but not everyone sees things the same way (e.g. [2]). Whatever, I
don't think there's actually such a rigid tool set to conflict with
RDF/XML, there are probably points where more interop can take place
(like the graph/XPath stuff) given enough imagination and time...

The tool that XML developers really, desperately seem to want to cling
on to is the text editor, and on that front perhaps Turtle is the best
solution.

Under what circumstances are XML developers likely to benefit from
RDF? Ok, moving from a tree to a graph has obvious benefits in the Web
environment. Inference and the logic of ontologies is pretty handy
too. Maybe the most immediate benefit is being able to mix together
data from different domains (or unambiguously extend within a domain).
How much of these can be done without committing to RDF/XML? Most, if
not all. But if a developer already has data in an XML syntax, what
are they to do?

I suspect there's still plenty of work that can be done (or broadcast
more, if it's already been done) in terms of using domain-specific XML
languages in RDF systems. Assume people don't want to add
rdf:parseType="Resource" every few elements (a fairly safe assumption,
IMHO). Domain-specific XML is usually relatively easy to consume, but
producing domain-specific material can be a pain. I think that pain
could probably be eased. So can't we perhaps leave commitment to an
RDF syntax until after the model has been introduced?

(btw - has anyone done a serializer that takes an RDF model together
with an XML Schema, filters down the statements to those which
correspond to names in the schema, then pumps out data that conforms
to the schema?)

I guess what I'm suggesting is that converting XML data to RDF is a
darn sight easier than converting XML developers to RDF. Once a decent
number of RDF-based systems are online (and joined together) the
utility should be obvious. Until we have widespread model envy,
integration through interfacing is probably the easiest path. This
kind of skirts around Chaals' excellent point, that they just want to
know what THE syntax is. Ok, tell them it's their current syntax.

So is promoting RDF+XML a lost cause? Definitely maybe. I certainly
don't think *using* it is a lost cause (as long as you don't look at
it too often). None of the alternate XML syntaxes seem to have gained
significant traction. For promotional purposes, where generalized
syntax must be addressed then Turtle almost certainly is the answer.
But like everyone's already said, it's the model that counts.

Cheers,
Danny.

[1] http://www.semanticplanet.com/2004/10/announcingSemanticPlanetsRDFLibAndCarp
[2] http://www.martinfowler.com/bliki/MovingAwayFromXslt.html

-- 

http://dannyayers.com

Received on Thursday, 25 November 2004 09:38:43 UTC