W3C home > Mailing lists > Public > www-webont-wg@w3.org > January 2003

RE: Review of Reference (syntax reference? no)

From: Smith, Michael K <michael.smith@eds.com>
Date: Fri, 3 Jan 2003 10:51:48 -0600
Message-ID: <B8E84F4D9F65D411803500508BE3221412C68B08@USPLM207>
To: Dan Connolly <connolly@w3.org>, Jeff Heflin <heflin@cse.lehigh.edu>
Cc: WebOnt <www-webont-wg@w3.org>

> >  One particular glaring ommission from the
> > document is that it does not define which elements need
> > parseType="Collection." For example, look at the Boolean combination of
> > class expressions section. There is not even a hint that you need to use
> > parseType="Collection" with intersectionOf and unionOf. The only way you
> > could guess this is if you look at the schema, determine that it takes a
> > List, and then use your RDF Guru knowledge to know that
> > parseType="Collection" is used as shorthand notation with Lists. 
> 
> Or read the guide.

Much as I like hearing those words, I thought the Reference in
combination with the formal sematics was definitive.  Which means it 
should nail down those things that are sentences in the three dialects.
Reference to some other document that contains a precise 
specification is probably ok, but it needs to be identified in the 
Reference.  

Perhaps this is why we were arguing 3 weeks ago about whether
a set of sentences was in the language or not.

- Mike

-----Original Message-----
From: Dan Connolly [mailto:connolly@w3.org]
Sent: Friday, January 03, 2003 10:03 AM
To: Jeff Heflin
Cc: WebOnt
Subject: Re: Review of Reference (syntax reference? no)



On Fri, 2003-01-03 at 09:36, Jeff Heflin wrote:
> If we rely on RDF/XML being the syntax, then at the very least we need
> to indicate which of the language "elements" are RDF properties and
> which are RDF classes, currently everything in Reference is called an
> element.

I agree completely. Get rid of "element" and use property
or class as appropriate instead.

>  We would also need, at a bare minimum, to give the domain and
> range for each property.

Yes, also a good idea.

>  One particular glaring ommission from the
> document is that it does not define which elements need
> parseType="Collection." For example, look at the Boolean combination of
> class expressions section. There is not even a hint that you need to use
> parseType="Collection" with intersectionOf and unionOf. The only way you
> could guess this is if you look at the schema, determine that it takes a
> List, and then use your RDF Guru knowledge to know that
> parseType="Collection" is used as shorthand notation with Lists. 

Or read the guide.

> However, that said, I think doing all of this would be more confusing
> than just simply laying out a grammar for the most commonly used OWL
> forms.

Er... the Guide gives examples of the most commonly used
OWL forms, no? It doesn't give a grammar, but is that
really necessary, or even useful?

>  I would think we want to make it as easy as possible for people
> to jump in and start using OWL. Many people may be turned off by the
> fact that they to read through RDF Core's 5 documents before they can
> even begin to understand OWL's 5 documents.

Such people are misinformed; they can start right in with
the OWL guide. And if they only want to learn enough
RDF to get by, they can just read the RDF primer.

Meanwhile, I'm pretty sure other communities (e.g.
implementors, logicians, etc.) are glad we didn't
limit ourselves to one perspective when specifying
these languages.

>  Furthermore, don't OWL Lite
> and OWL DL have syntactic restrictions not present in RDF? If so, we
> must make these explicit and not just simply say look at RDF for the
> syntax.

They're explicit in the semantics doc; what users really need
to know is in the guide.

I'm starting to wonder about the purpose/necessity of the Reference
document. It certainly played a critical role in getting
to this point, but I wonder about the value of maintaining
it going forward.

Here are some purposes that folks seem to think
it serves, but I don't:

  * it's the place to start if you want to know about OWL.
	No, I'd recommend the feature synopsis for that.
	I like the way the discussion of the three
	dialects is being elaborated in the feature synopsis;
	I think the requests for a "roadmap" to
	the other documents might usefully be addressed
	by adding a section to the feature synopsis.

  * it's the definitive specification of the language.
	No, it's actually not exhaustive. The semantics
	doc is, short of one or two things:
	 -- application/owl+xml media type
	 -- "The exchange syntax for OWL is RDF/XML, as
	specified in the OWL Reference Description [OWL Reference]."
		if we just change that to directly cite the
		RDF syntax spec, we're golden. RDF concepts
		is already cited (in order to import the
		definition of RDF graph)

  * it's useful reference documentation for users and implementors.
	Hmm... really? I think implementors are going to
	read the synopsis and maybe a little bit of the guide,
	then code 'till they think they're 80% done; then,
	if they're serious, they'll start griding away
	at the test cases, and when they get unexpected results,
	they'll be referred to the semantics doc and/or guide,
	where they'll learn about the other 80%. 1/2 ;-)

	Users will use the guide. Successive elaboration
	is much more useful than exhaustive enumeration;
	but if it's exhaustive enumeration you're after,
	the "Term Index and Cross Reference" section
	is there just for that.
http://www.w3.org/2001/sw/WebOnt/guide-src/Guide.html#TermIndex

  * it's the only place where media types are discussed.
	Well, OK, but we could use a much shorter document
	to specify media types for OWL.

  * it's where the presentation syntaxes are discussed.
	These are better off in their own documents,
	if you ask me. But perhaps they could
	fit in with the owl media type spec.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Friday, 3 January 2003 11:51:58 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:56 GMT