W3C home > Mailing lists > Public > www-rdf-comments@w3.org > April to June 2002

General Comments about RDF Primer, Model Theory, and Revised Syntax WDs

From: Bijan Parsia <bparsia@email.unc.edu>
Date: Fri, 07 Jun 2002 15:24:02 -0400
To: www-rdf-comments@w3.org
Message-ID: <1261225632.1023463442@localhost>

Thanks to Dave Beckett for proding me to do this.

Detailed, nitty, comments to follow separately, at least about the MT and 
RS WDs. Here, I'm focusing on more general issues especially about how the 
Drafts mesh together.

Before getting into the qualms and quibbles let me just say that the array 
of WDs, even in their current unfinished state, leave me feeling a whole 
lot better than RDFM&S ever did. I commend the working group and especially 
the respective editors for their hard and fine work. The light at the end 
of this Rec tunnel seems bright and pleasing.

The most important overall issue in my book is that the three specs 
(Primer, MT, and RS) don't yet work as one. There are some inconsistencies 
(for example, the RS talks of anonymous nodes as nodes with "unknown" URIs, 
whereas the MT denies that they have URIs of any sort), but there is also 
quite a lot of stylistic and presentational diversity. For example, the big 
one is:

VARIENT LINEAR TRIPLE SYNTAX

Primer uses faux N-Triples (illegit line breaks). MT uses N-Triples + [] to 
force qname expansion (not to mention a funky faux uri scheme). RS uses 
N-Triples (albeit correct ones for their mandated purpose). Whatever is 
decided upon, I really hope that *gratuitious* variation is avoided. I 
would suggest, too, that the MT use N3 syntax rather than the invented 
bracket one, to avoid yet more proliferation of triple syntaxes. (I don't 
see that adding square brackets adds to the clarity of the triples.)

Actually, this is one of those times when I wish we had a standard way to 
let the user configure the presentation syntax. You want N-Triples? Fine. 
You want only pictures? Sure! You want F-Logic? Go ahead. System 
documentation could include the specs customized to their own preferred 
syntax.

Other general comments:

UNSERIALIZABLE GRAPHS

Both the MT and the RS talk about RDF (well-formed RDF?) graphs that can't 
be serialized by RDF/XML (though the wording in the MT suggests that there 
are graphs that can be serialized by N-Triples. I sorta assume that the 
missing graphs aren't a big deal for most current apps, but it would be 
nice to have a bit more explicit discussion of this in the RS--and if it 
*is* a big deal in the Primer as well. I'm pretty sure I know what causes 
problems, but this is something that should be dealt with up front.

RDFS INTEGRATION

RDFS's model theory is part of the RDF MT document, RDFS is extensively 
discussed in the Primer, yet there is a separate RDFS document which sorta 
addresses RDFS syntax (and discusses informally the semantics), but the RDF 
RS doesn't mention it at all. What's being aimed at here? Are RDF and RDFS 
tightly or loosely coupled, at least from the POV of the W3C? Do you want 
to encourage most RDF implementers to also support RDFS? I think 
encouraging RDFS level functionality is a good idea, *but* if it *is* going 
to be treated as something more or less completely different, then I think 
it should be better presented that way. E.g., move the Primer and MT bits 
discussing RDFS into the RDFS draft (or otherwise factor them out). 
Actually, I can see a case for some discussion of RDFS even in a purified 
Primer but mostly as a pointer to a richer document. Hmm. Maybe a separate 
RDFS + WebOnt primer would be best.

PRIMER'S JOB AND AUDIENCE

I'm unclear what specific task the primer sets out to do, and what its 
intended audience is. There are bits of the other drafts that seem better 
suited to the primer (e.g., RS section 2), and bits of the Primer that seem 
to belong elsewhere.

I'm not sure that the "guide to all the specs" tact is the most useful for 
the Primer (does it really have to talk about the *test case* document!?!?) 
For people who just want to use RDF and RDF based tech, it seems that 
there's an overwhelming amount of info in this doc that they probably won't 
need (RDFS? MT? Test cases? Example vocabularies?) and crucial missing info 
to getting started (what URIs can you/should you use to name stuff? how do 
you embedded RDF in 1) HTML, 2) SVG, 3) other formats, etc.).

Indeed, I'm very skeptical that I could hand the Primer as it's now set up 
to a class full of library science students and have them come out of 
reading having the first clue about what to *do* next. (Actually, I think 
RDF in SVG is a pretty killer, compelling example, especially with 
co-depiction like demos.) I mean, what can I do with RDF *today*?

If it's primarily for implementers, tech experts, and the like, then it's 
got too much wifty, wooley stuff surrounding the core info. Send these 
folks to the actual specs faster :)

If it's for managerial sorts who direct adoptation of technology without 
really understanding it, there isn't *enough* of the wifty, wooley stuff!

The self-proclaimed job:

"""This Primer is intended to augment the other parts of the RDF 
specification, to help information system designers and application 
developers understand the features of RDF, and how to use them."""

Seems rather a big, ill-defined chunk to chew.

Cheers,
Bijan Parsia.
Received on Friday, 7 June 2002 15:22:20 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 21 September 2012 14:16:30 GMT