Re: XML Serialization

> > <digression type="paranoid">
>...
> > </digression>
>
>Cute, but I think it's just a case of people try to solve a different
>problem.  XML is reasonable for marking up texts

Yes. The problem is that its become used as a way to write arbitrary 
trees, and for that it sucks, but not quite badly enough to be 
obvious to everyone.

>-- like indicating
>how a paper should be formatted -- isn't it?   The redundant end tag
>is a bit silly, but probably good human engineering, like commenting
>your #endifs.   (And of course, you used it brilliantly in your
>e-mail.  :-)
>
>Now, why did the RDF WG chose XML instead of s-expressions or
>something else elegant?  I wasn't there, but I love rumor mongering
>and wild speculation.  Maybe they figured in the mood of the day, it
>would give RDF a leg up.  And it probably did, with the librarians.
>Perhaps it wasn more of a leg iron to the computer scientists, though.

Quite, and elegantly put.

> > A constructive suggestion: Why not make N3 the official version of the
> > RDF syntax?  There would be several advantages:
> >
> > > It would eliminate confusion over the sense in which RDF graphs are
> >   a notation, not a domain of discourse.
> >
> > > It would be serialized *already.*  Because N3 is the primary
> >   serialization, the XML serialization would be secondary, as it
> >   should be.
> >
> > > N3 embodies a goodie or two don't exist in the official RDF, and
> >   this would be a good way to sneak them in.
>
>While I heartily agree with your gist here, I have a technical reason
>why not: N3 has lots of issues, too.
>
>I do wish we could have the debate over what is the right language,
>but I think the only practical way to make it happen and make it
>involve little loss of blood is to be secure in the knowledge that the
>language doesn't really matter, because it's the graph that counts.

Right. Keep that in your sights. It's the graph that counts. Take a 
look at CG graphs, for example: all of logic in a nutshell.

>If we do pick a character-sequence notation as the fundamental
>standard, I think it should be as simple as possible.  Something like
>N-Triples, although I can think of at least four big issues with even
>something that simple.  (identifier syntax, literal syntax [and type],
>equality, and nesting.)

Nah, the main (only?) N-triples issues are re-naming anonymous node 
labels and how to ignore triple ordering. All the others are issues, 
but they are issues in the RDF graph itself. RDF can't do nesting, 
for example. It just can't; nothing to do with Ntriples. (RDF++ maybe 
will, one day, but....)

Pat Hayes

---------------------------------------------------------------------
(650)859 6569 w
(650)494 3973 h (until September)
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Thursday, 30 August 2001 19:48:11 UTC