W3C home > Mailing lists > Public > www-rdf-interest@w3.org > November 2004

Re: Is promoting RDF+XML a lost cause?

From: Phil Dawes <pdawes@users.sourceforge.net>
Date: Thu, 25 Nov 2004 17:09:16 +0000
Message-ID: <16806.4540.50498.171411@gargle.gargle.HOWL>
To: Giovanni Tummarello <giovanni@wup.it>
Cc: www-rdf-interest@w3.org

Hi Giovanni,

Giovanni Tummarello writes:
 > 
 > My2c :-) ..
 > 
 > it is *looking* at the XML in the RDF/XML serialization that can be 
 > considered a lost cause. People should'nt  look directly at it.
 > Its like if when the JPEG format was invented instead people said its a 
 > lost cause since if you look at it with a hex viewer you dont see much
 > .. and we should all use ascii art instead.

I'm not sure this is a good analogy - jpeg is a binary format, and is
for most intents and purposes 'unstructured' data. (Although
technologies like face recognition and OCI help to mine structure from
it).

I think the main reason that XML is popular and successful is that
people *can* look at it, and more to the point, the structure maps
intuitively to the logical model of the contained data (i.e. a tree
with attributes) - this makes it easy to understand, and removes a
layer of abstraction (the mapping of the serialization to the model),
at least at a conceptual level.

(I should also note that one of the reasons XML is so popular is that
quite a few people had already got used to looking at triangle
brackets with HTML, so it already felt 'familiar' to a large audience.)

 > RDFXML does ok the serialization problem, i can export from jena and 
 > import it in sesame... everything else needs to be solved at a different 
 > level

True, but how many RSS sources serialize XML from an infoset model
using e.g. a DOM? None that I know of - they use a hand-written
template to serialize the data into XML. Sometimes they just print it.

In fact at work most of our RDF/XML is generated using this sort of
technique - for a quick hack glue job when you haven't got the time or
patience to install and learn an RDF library, nothing beats 'print'ing
the XML out. This is not necessarily a good idea, but it is a common
approach and makes XML an easy technology to promote (very low barrier
to entry).
 
 > in fact.. i believe that in order to  widen acceptance, 
 > people shouldnt be made to approach RDF in a way so tangled with XML as it is
 > in the RDF primer.

Agreed, but to compete with XML in terms of popularity I think there
needs to be an easy serialization that maps cleanly and conceptually
to the RDF model. A subset of turtle is it IMHO, but I don't know how
well a non-XML serialization will go down with XML-committed
developers.

 > It's the model and the semantics that matter and make rdf more
 > powerful and actually simple

True, but people care about serialization, especially if they're
writing it out using 'print' statements.

 > It's a graph .. so no textual serialization will ever make it clear? 

Maybe, but it's also triples which is easy to make clear. I think
turtle/n3 does a reasonable job.

(actually I tend to think triples rather than graphs, but that may be
because I've spent time developing a triplestore)

Cheers,

Phil
Received on Friday, 26 November 2004 10:34:25 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 5 February 2014 07:14:57 UTC