Why RDF was a good choice

Hello,

I have only recently started looking at CC/PP for my research, and therefore
I was not involved in the development of the CC/PP language. However, I feel
I should defend the choice for RDF as a representation language for CC/PP.

As stated by Tayeb and others the last few days, the benefit of using RDF is
in the semantic advantages provided by this language. A related benefit is
the possibility to cooperate with other metadata languages. In particular, I
am thinking about the MPEG-7 Multimedia Content Description Interface, even
though MPEG-7 is not RDF-based.

In my opinion, the problems on using RDF have one major cause: incorrect use
of RDF (I believe this is stated by Mark in one of his recent papers).

The major problem is probably the complicated syntax for the XML
serialisation of RDF. One way to solve this is to use some sort of
RDF-editor (not just an XML editor), when composing a device profile. But in
my opinion, it would be a good idea not to limit the serialisation of the
RDF graph to XML, but also allow the N3 notation. Perhaps it might even be
useful to develop a proprietary notation specifically for CC/PP? The major
requirement for such a notation is that the original RDF graph can be
(re)constructed in an unambiguous way. Note that backwards compatibility
with UAProf remains as long as the XML notiation is allowed.

I am also in favor of a validation mechanism for client profiles, as has
been suggested in this mailing list. But I also want to state some remarks.
Firstly, RDF should not just be used as an underlying datastructure, because
this limits the semantic advantages of the language. In my opinion, the
value of RDF data is seriously increased by using RDF Schema or a layer on
top of RDFS, such as DAML+OIL. In the case of CC/PP, new vocabularies should
extensively use "rdfs:subProperty" statements, as this enables forward and
backward compatibility between vocabularies. For example, a server that
supports the old vocabulary X with property A, can still understand property
B of a client profile, even when this property belongs to a more recent
vocabulary Y, as long as the server can somehow learn that "B is a
subProperty of A". Therefore, the expression and the distribution of CC/PP
vocabularies is very important for the practical use of CC/PP. Developing a
core vocabulary as part of the CC/PP standard might also be a good idea.

As a conclusion, I repeat that RDF is a good choice for CC/PP, as long as it
is used in a correct way, and is added by the use of RDFS or a layer on top
of RDFS.

I hope I made myself clear in this mail, and I welcome all comments.

Greetings,

Sam Lerouge
Department of Electronics and Information Systems - Multimedia Lab
Ghent University, Belgium
sam.lerouge@rug.ac.be

Received on Thursday, 6 June 2002 10:18:18 UTC