Re: missing bit of RDF for XML people

On Thu, 03 Feb 2005 10:29:04 -0500, Stefano Mazzocchi
<stefano@apache.org> wrote:

> Roy is right in one thing: making Atom in RDF/XML would make it a lot
> harder to market it against RSS (not 1.0, the XML ones). It would be
> *way* easier for everybody if people stopped arguing about "should be
> RDF, no it shouldn't" and thought about ways to build a bridge so that
> data can move from one island to the other.

Hmm, it's academic now, but the arguments against using RDF/XML for
syndication have been primarily either due to the complexity of the
syntax, or the complexity brought in through buying into the RDF
model. The former point is difficult to counter when comparing RSS 1.0
(slightly constrained RDF/XML) to Userland RSS 0.91 (plain XML,
without namespaces). But Atom is a far more complicated syntax than
RSS/XML (for a start it uses namespaces). Regarding the complexity of
the model - it's not a clear case either way. RDF *looks* complicated.
The domain model is relatively simple, but lacks any formal semantics.
This is fine until you try mixing in data that isn't expressed in the
core. Simple things are very simple, complex things are a lot harder
than they need to be.

> As I said previously, having a 'semanticsheet' would be enough to make
> the Atom people happy with their less verbose and more palatable XML and
> the SemWeb people happy because they can "augment" that the implicit
> encoding with the semanticsheets... and, hear hear, if that "obvious"
> interpretation drifts overtime (or different groups interpret it
> differently), it's just a matter of changing the semanticsheet and not
> the entire pile of data that the world already generated.

Yep, that's essentially the way things have gone, the 'semanticsheet'
primarily being XSLT, although I'm hoping the model will be tied down
a little too.

> [Here Roy is dead wrong: there is no such a thing as an "obvious
> understanding" of a syntax... even of RDF/XML]

I believe he meant of anything-but RDF/XML, e.g.
<title>The title</title>
Where the obvious falls down is when you try and derive some kind of
machine-readable interpretation - title of what?

> Having RDFizers is a *lot* more useful than having a pile of explicitly
> semanticized data... RDF semantic is explicit, but no semantics is
> carved in stone: another layer of indirection is actually a plus, not a
> minus. It avoids you to spend a few teraflops in OWL inferencing that
> can be done with a few minutes of XSLT.

Hmm, I'm not entirely convinced for the general case (cliches like
"premature optimization" spring to mind), although certainly in a lot
of cases a little XSLT can save a lot of cycles.

> But of course, the "thou shall use RDF" verb has been ringing so loud
> and so obnoxiously over the years that either people take it for granted
> or hate it just because... me wonders: what happen to those who just
> wanted to solve the problems instead of marketing visions?

Again - hmmm. I was personally attracted to RDF in the first place
because it offered a way of handling some practical stuff (loosely
structured catalog/search) that at the time was more difficult by
other means. Never looked back...well ok, I still use a lot of the
same tech as I did then, only now there are at least alternatives ;-)

Cheers,
Danny.

-- 

http://dannyayers.com

Received on Thursday, 3 February 2005 17:03:39 UTC