Comments on datatyping draft

Ref:  http://www.coginst.uwf.edu/~phayes/RDF_Datatyping060102_draft.html

Generally, I have a niggling concern that some of the issues here are 
sufficiently subtle that they might cause confusion among developers.  In 
particular, the operation of rdfd:dcv seems tricky:  I had previously come 
to understand it (or it's predecessor) as operating pretty much at a 
syntactic level by restricting lexical forms.  Now it's not so clear.  The 
model theory (section 5, items (4,5)) seems to impose a purely syntactic 
constraint on "LLL" in terms of L2V, but some of the text elsewhere seems 
to focus more on values (e.g. section 3.4.1, and even the derivation of 
rdfd:dcv, seem to be focused on values).

Alternatively might the datatype be considered to relate to lexical values 
rather than literals;  e.g.

    Jenny age "10" .

meaning the same as something like:

    Jenny age SomeAge .
    SomeAge xsd:string "10" .

where xsd:string is assumed to be a datatype property that relates strings 
to themselves;  i.e. it is an identity relation on lexical values.

Otherwise the broad thrust looks fine to me, within the constraints of tidy 
literals.

...

Some detailed comments:


Abstract

Considering that this document is explicitly intended to be read as an 
extension to the existing   RDF specs, I think the first two sentences of 
para 1 are redundant, and could be deleted without loss of useful content.


Section 1

For similar reasons, I think that all of the 1st para, and all but the 
first sentence of 2nd para, could be deleted.  Also, delete "this" from 
Para 2, 1st sentence.


Section 1.1, para 3

On first reading, the term "datatype context" felt like a reference to a 
new, undefined concept - which maybe it is.  Would it be sufficient to just 
say "association... with a datatype." at this point.  Also, could "denoted 
explicitly by a blank node" not equally be "denoted explicitly by a node" - 
I think the node concerned here could quite reasonably be labelled with a URI.


Section 1.2

I don't think any introductory verbiage is particularly needed here.

8th and 9th bullets refer to "RDF schema".  This use of "RDF schema" seems 
to relate to RDF schema vocabulary which I think is not the 
intent.  Rather, I think the intent is to refer to documents, which might 
be separate from the main document, that contain information about the 
properties and classes applicable to all of their uses in the main 
graph.  (I've encountered a similar issue explaining some related issues 
concerning CC/PP:  I think RDF Schema should be regarded as just an initial 
vocabulary of things one might say about the properties and classes used in 
an RDF graph - it's sometimes convenient to talk about an RDF schema 
document meaning a document that contains such assertions, without 
necessarily being limited to RDF Schema vocabulary.  Indeed, in this case, 
the intent is to use RDF Datatyping vocabulary in this way.

Hence, I suggest replacing "RDF schema assertions" with "assertions about 
RDF properties and classes".

I think the final paragraph of this section could be dropped without loss 
of useful content.


Section 2.1

Here, and later, there is mention of "canonical datatype mapping" but, as 
far as I can see, such canonical mapping plays no part in the datatyping 
scheme.  I suggest dropping such references;  i.e. delete 3rd para.


Section 2.2, final para:

Suggest:
"RDF Datatyping only provides ..." -> "RDF Datatyping provides only ..."


Section 3, 1st para:

The final sentence says "Global datatyping requires smaller graphs...", 
which I found confusing.  I think the intent is: "Global datatyping allows 
smaller graphs..."


Sections 3.x:

When I print this document, the tables containing RDF/XML and graph 
diagrams are too wide to fit on the page.  I suggest placing them one above 
the other in each case.

Also, in a number of these tables, the text does not format properly -- 
text is overprinted and unreadable:
- Section 3.1, 1st table
- Section 3.4.1, 1st table
- Section 3.4.1, 3rd table
- Section 3.4.2, 1st table


Section 3.1, 3rd para:

This paragraph talks about nodes "indicating" values:  is "denoting" meant 
here, and if so is there a good reason not to use that term?


Section 3.1, 4th para:

"would always be invalid" -- I had come to understand that "valid" means 
true in all interpretations, so I'm not sure what "always be" imparts here.


Section 3.1, 6th para:

The phrase "restricted entirely to the particular statement" caused me to 
stumble -- I can't see any significance to the "entirely" here:  is it needed?


Section 3.1, 8th para:

"It is the pairing of the lexical form and datatype together which 
determines the particular value..." I think is not right - I don't see that 
the pairing of itself determines anything.  I suggest "The particular value 
is determined by its datatype relationship with the lexical form, ..."


Section 3.3, para 2:

I didn't understand the phrase "...describing datatype values as property 
values, ..."


Section 4.2, final para:

Missing capitalization at start?


Section 6, class rdfd:Datatype:

suggest reference to canonical lexical space could be removed.


Section 6, property rdfd:dcv:

The comment here really made me wonder whether rdfd:dcv was intended to 
operate at a syntactic level on literals, or at a value level on things 
denoted by literals.  The text here suggests to me the latter, but the 
model theory suggests the former.


...

That's it.

My apologies if some of these comments are a bit terse -- it's late and I 
wanted to get them out before I lost the impetus.

#g



-------------------
Graham Klyne
<GK@NineByNine.org>

Received on Wednesday, 5 June 2002 18:54:37 UTC