Re: 2002-04-16 RDF Datatyping WD submitted for review by WG

On 2002-04-17 13:31, "ext Dave Beckett" <dave.beckett@bristol.ac.uk> wrote:

>>>> Patrick Stickler said:
>> 
>> The first complete draft (excluding introductory and non-normative
>> material) is available at
>> 
>>    http://www-nrc.nokia.com/sw/rdf-datatyping.html
>> 
>> and is presented to the WG for review.
> 
> I reviewed the above since the
> http://lists.w3.org/Archives/Public/www-archive/2002Apr/0039.html
> 
> was zipped up and not possible to read online.

It hasn't changed.

> 
> Summary:
> 
> Looking good, I'm happy to see it published if the usual w3c doc
> production stuff is completed.
> (I won't be at Friday telecon)
> 
> I'd like to compare it to Pat's simple datatyping 2 now that it is
> back online but I've not got the time this week.
> 
> Detailed Comments:
> 
> 1.4
> 
> Are the rdfd namespace uRIs reserved for your use?
> I.e is it OK to use http://www.w3.org/2002/rdf-datatyping

As far as I know.

> Maybe ask a W3C team person to reserve/create some URI to use
> for this.  

I was thinking we'd wait until we decided on the whole namespace,
version issue. 
 
> Have you got an RDF Schema for the terms you define?

Keep reading ;-)

It's in section 6.

> 2.1.1
> rdfd:Datatype
> 
> Not sure what you are defining here - a term "rdf:Datatype" or
> something in an XML namespace.
> 
> I expect the former, just for clarity?

I'm actually not sure what your asking. rdfd:Datatype is
an RDF Class that exibits the characteristics defined for
RDF Datatypes.

Instances of rdfd:Datatype may also be datatype properties.

> However reads confusing since in the first examples
> it says "an rdfd:Datatype ..." then goes on to show some syntax
> with rdf:* things and no sign of rdfd:Datatype.

I tried to keep the examples minimal, including only the
statements necessary.

Can you elaborate on how it was unclear or confusing, so I
can better understand how it could be improved?

> Maybe also mention what the rdfd: namespace is for, why it is
> separate from other namespaces?

If the namespace stays, definitely.

> 
> 3.1
> 
> examples need numbers, captions and links
> 
> Illegal examples need distinguishing from legal ones

Good suggestions. Thanks.

> (Aside: The contracted
> 
> <rdf:Description rdf:about="#John">
>  <ex:age xsd:integer="25"/>
> </rdf:Description>
> 
> form can *always* be used since the lexical form of
> a datatype is always a string.  The longer form is still useful
> to match the graph pattern (node, arc, node, arc, node)
> and help explain it.
> )

Right. I wanted to show both forms for the sake of clarity. Most
folks will probably want to use the contracted forms, but the
expanded forms better show the relation of the XML to the graph.

> End of 3.1:
> "Obviously, this is only valid when the literals do in fact map to
> the same value under the respective datatype mappings."
> which is as you mentioned above, for RDF datatyping applications
> to determine, not RDF.

Right. I should add that clarification.

> 3.2
> 
> (Introduces rdfd:lex as an RDF Property.  Which makes the rdfd:Datatype
> earlier look even odder.)

Well, rdfd:lex *is* an RDF Property. But it is not a member of
the class rdfd:Datatype.

Because of the datatype property idiom, members of rdfd:Datatype
*are* odd. That's just the way things are.

Is the nature of either rdfd:Datatype or rdfd:lex unclear, or
is it just that there is oddness?

> 3.4.1 rdfd:datatype
> 
> is this different from rdfd:Datatype?

Quite.

> If it is the same, then
> you need a forward pointer from the first mention.  If not, it
> is confusing.

The rdfd:datatype property associates an rdfd:Datatype with
a general RDF property.

> 4.2
> 
> Link to the RDFD MT since you reference it.

Right.

> 4.3 "versus"
> 
> maybe a bad choice of phrase. Something like "Constraining values with
> rdfd:datatype or rdfs:range"?

Well, the section is comparing RDF Datatyping mechanisms with
those of RDF Schema, and the choice is between rdfd:datatype
and rdfs:range, so the term 'versus' seems correct.

> Although Dan Brickley has been backing away from using the constraint
> word with RDFS

It's a matter of perspective. Perhaps it's more accurate to say that
one is constraining the datatyping interpretation, such that invalid
lexical forms or undefined datatype contexts have no interpretation.

> Give examples of legal/illegal uses according to the constraints
> with either/both of rdfd:datatype and rdfs:range.

I thought I did ;-)

> 4.4
> 
> Needs at least 1 example

Agreed.

> spelling: eqivalent

Thanks.

> 4.5
> 
> Needs at least 1 example

Agreed.

> 5
> 
> Link to definitions in RDF MT that you use

Right.

> 6
> 
> Ahah, the answers to my earlier questions.  But could have been explained
> earlier.  And I think it was more being used as a shorthand for "an RDF
> datatype interpretation as defined by this specification"

Right. I can put a mention of the RDF Schema up front for folks who don't
bother to look at the TOC ;-)

> Hmm.  This doesn't make sense:
> 
>  <rdfs:Class rdf:about="&rdfd;Datatype">
>     <rdfs:label xml:lang="en">RDF Datatype (Property)</rdfs:label>
>     ...
>     <rdfs:subClassOf rdf:resource="&rdf;Property"/>
>  </rdfs:Class>
> 
> 
> It is an rdfs:Class and an rdfs:Property.  Surely that label is confusing.

The class is not a property. But its instances are.

It's reasonable for the label of rdf:Property to be "Property". Likewise
it's reasonable for the label of rdfd:Datatype also to indicate that
it is a Property class.

It's the duality of rdfd:Datatype that's odd and confusing, not the
label. The label is quite accurate.

That's what you get from the datatype property idiom.

> Not sure of the RDFS style of:
>     <rdfs:domain rdf:resource="&rdf;Resource"/>
> 
> Everything is a resource right?

That remains an open question for me.

Unless and until I see something like

   rdfs:Literal rdfs:subClassOf rdf:Resource .

in the RDF Schema spec, I consider the above to be correct.

Dan?

> You don't have range/domains for the rdfd:Datatype saying that
> they take resources(literals)

How could I? rdfd:Datatype is not a property, therefore has no
range or domain itself.

However, the closure rules defined in section 5.1 provide the
fixed domain and range for instances of the rdfd:Datatype
class. The domain is defined explicitly by rule 1a and the
range is inherited from rdfd:lex by rule 1b.

So that's covered.

> 7.1.2
> 
> a "non-literal resource" !
> 
> Not sure that phrase helps much

Feel free to suggest an alternative. That's the best I could
think of ;-)

> 7.2.3
> 
> The Dublin Core doc for this is
> http://dublincore.org/documents/2001/11/30/dcq-rdf-xml/
> 
> (latest version http://dublincore.org/documents/dcq-rdf-xml/ )
> 
> which is a DCMI proposed recommendation.
> 
> 
> The URL Aaron gave is his own work, turning it into Ntriples and
> has no status with the DCMI.

I'll let Aaron respond to that ;-)

> General
> 
> Maybe have ntriples output too; noting that the datatyping idioms
> don't introduce any new triples.

Have them where? In section 7.1? Everywhere in the WD?

> Link to test cases / entailments?
> 
> Need more links throughout to the cited references such as RDF Schema
> model theory, maybe to sections of the latter?
> 
> The pictures are all(?) JPGs and look rather fuzzy on my desktop around
> the words; can you turn them into PNGs with solid colours?  Maybe this
> is anti-aliasing?  The picture in section 4 is especially bad.
> 
> When you use phrases like "the inline datatyping idiom" make that
> a link to the section that defines it.

These are all very good comments, thanks.

As I said, I'm focusing at the moment on the technical details and will
be doing alot of polishing and whatnot once that is nailed down.

> References
> 
> Lots of w3c doc style things to fix :)

I'm sure ;-)

Thanks for your comments,

Patrick 

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Wednesday, 17 April 2002 07:27:08 UTC