W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > July 2003

Review of RDF semantics editor's draft July 15 2003

From: Dave Beckett <dave.beckett@bristol.ac.uk>
Date: Thu, 24 Jul 2003 12:20:57 +0100
To: w3c-rdfcore-wg@w3.org
Message-Id: <20030724122057.235f1bf3.dave.beckett@bristol.ac.uk>

I reviewed:
  http://www.w3.org/2001/sw/RDFCore/TR/WD-rdf-mt-20030117/
  Last Modified: 07/18/03 16:55:56
  221233 bytes


Summary
  Having read through in detail, I'm happy with this document.


Although I am not a model theoretician, this is comprehensible to me
and I feel the changes reflect fairly the updates to the LC version
in what might be called substantial editorial changes along with
addressing the last call and post-last-call issues and decisions.

As such, in my opinion, it seems suitable for proceeding along the
recommendation track rather than as another last call draft.


Some points:

* I'm not so clear on the exact correctness of the axiomatic triples,
  I'd need somebody with code to check these.

 * Similarly with the 7.x entailment rules I'd want other eyes or
   running code for them.

 * Not reviewed: mapping to lbase or proofs of lemmas appendices.
   I would be happy if these were either made informative or dropped.


-----
Detail

Change List
This should go to a Changes section appendix and ordered somehow - by
section changed or issue (with pointers into the changed sections).


Globally

I think the advice I saw on style means that URIref should
be spelt out URI reference everywhere.

simple literal=>plain literal

You don't need to write in the third person ("the editor").


Abstract
OK

Status
Not reviewed - this will change.
However the status will need to carefully reflect the changes.

Table of Contents
You should label any informative appendixs here too with "(Informative)"
Are appendixes A & B part of the document?


0. Introduction
0.1 Specifying a formal semantics: scope and limitations

P1
Recommend w3c style:
"see the RDF Concepts document [RDF-CONCEPTS]"
=>
"see _RDF Concepts_ [RDF-CONCEPTS]" where _RDF Concepts_
is a direct link to the concepts doc

P2
CSS style.  I note 'semantics' is a glossary word and linked
only via mouseover.  It would be nice if this was clear visually

P3
All about model theory till the last sentence:
  "This provides the maximal freedom for implementations while
  preserving a globally coherent notion of meaning which allows RDF
  to be used to transmit content."

I think that would be better split.
  "This provides the maximal freedom for implementations while
  preserving a globally coherent notion of meaning.  
  These features of model theory are appropriate for specifying RDF's
  semantics in transmitting content."

or something like that.  It might even be better moved to a later
paragraph since P4 continues describing model theory.

The 'vocabulary entailment' links do not go to anything.


P4
unchanged since last call

P5
"This document gives two versions of the same semantic theory:
directly, and also (in an informative appendix)"

directly - also might benefit from adding "(normative)" to
counterpoint the informative.

link to the appendix, give it's title.

The red text might be better given as an indented (Normative) Note.

P6-P8
unchanged/minor word changes since last call.

  P7 - DAML+OIL not DAML?  OWL - which document are you citing, give
  it's full title.

  P7 RDF graphs - should this not link to the definitions in RDF Concepts?


0.2 Graph Syntax

The editorial changes are OK


0.3 Graph Definitions

P1
"RDF Graph" points to graph syntax in concepts, not the definition
term: http://www.w3.org/TR/rdf-concepts/#dfn-rdf-graph (as used
in 0.2 P1)

P4
"fixed by the RDF semantic rules"
=> please link or cite to this

"typed literal containing a URIref in V"
presumably also explained elsewhere, please link to the (later) defn.
The last sentence also describes this partially.

Maybe replace 
  "A name is from a vocabulary V if it is in V or is a typed literal
  containing a URIref in V." 
with
  "A name is from a vocabulary V if it is a URIref in V or is a typed
  literal containing a URIref in V." 
for consistency?

Maybe:
"The names of a graph are all the names which occur in the graph."
=>
"The names of a graph is a set of all the names which occur in the graph."

since the next sentence says it is a set.

P5
"set of names" should be italic not red

P6-P7
OK

P8
 "A proper instance ...
  or two blank nodes in the graph have been identified."

I know what you mean here but that last clause doesn't work well
for me.  You want to say that two blank nodes have been "merged"
I think identified isn't suffient.

P9-P14
Ok

P13
"<ex:a> <ex:p> _:x."
=>
"<ex:a> <ex:p> _:x ."
i.e. separate the final term from the '.'

P15
OK

P16
"see the RDF syntax specification [RDF-SYNTAX]."

"see _RDF/XML Syntax (Revised)_ [RDF-SYNTAX]."
and link to the document 


1 Interpretations

1.1 Technical Notes

This is a very generic title - Why are they technical or Notes?
Maybe something like:
  Model theory technical approach
  MT Restrictions and extensions
?

P1-P4
OK

  P2 - Intensional glossary term link is wrong - it goes to #glossMonotonic


1.2 Urirefs, Resources and Literals.

P1-P3
  OK

 P3 'entity' is called out as a special word, but has no more meaning
 to me than "thing" or "thingy".  Maybe that's OK as a familiar term
 for anything?

P4
  Remove the red colour.
  Link "section 3.4" to the section.

P5
  "Urirefs"=>"URIrefs" or whatever.

  You are using "simple literals" here.  Earlier in 0.3, you used
  plain literal.  The correct term from RDF concepts is plain.

  The exclusive canonical XML link is broken here and in the
  references. Should be http://www.w3.org/TR/2002/REC-xml-exc-c14n-20020718/

  "exclusive Canonical XML" => "_Exclusive XML Canonicalization_"
  i.e. with a link, wrong title

  "RDF concepts and abstract syntax document[RDF-CONCEPTS]" =>
   "_RDF Concepts And Abstract Syntax_ [RDF-CONCEPTS]"
   i.e. with a link, capitalised


1.3 Interpretations

P1-P3
  OK

P4
  not "postponed until later sections" but
  "described in sections X.Y" with a link

P5
  Since you have several types of tables coming up, I don't think
  colour is sufficient to differentiate them (accessibility - use of
  red & green particularly bad!).  Add titles to the top such as for
  the next table P6 as suggested below.

  More red text, needing removal and adding links to the sections.

P6 (table)
  Give it a title or some way to refer to it (maybe a number)
  "Semantic condition 1 : Simple interpretation" ?

  Using 'literal character strings' however earlier you only defined
  'character strings' (1.2 P4)

P7
 OK

P8
 OK.  Maybe add a forward pointer to when datatype info is inconsistent.
 
P9
 OK

1.4 Denotations of Ground Graphs

P1-P3
  OK

P4
  unchanged since LC

P5-P6
  OK, minor typos changed since LC

  Maybe move "IP={1}" to a line on it's own.
  It seems odd to have it at the end of the IR line (although both
  are sets)

P7-P10 ("For example")
  unchanged since LC

P11
  spelling: "datyped interpretations"
  and add a link to where this will be ruled out.

P12-P18
  OK

P19
  In red again.  I'd remove "the fact".
  grammar: "in [the] predicate position"


1.5. Blank Nodes as Existential Variables

P1
  unchanged
  please link to the appendix with the proof

P2
  OK, minor rewordings    

P3 (semantic condition)
  unchanged

P4
  OK, minor rewordings

P5-P10 ("Note that")
  unchanged

P11 (new para)
  Merging was defined earlier formally, maybe link in here?
  Maybe link forward from there since blank nodes & merging is
  important to know about.


2. Simple Entailment between RDF graphs

P1
  minor changes OK

  grammar: "and a set S of (simply) entails a graph E if "
    probably should be "and a set S (simply) entails a graph E if "

P2-P5
  minor changes, OK

  P5: "the appendix" => replace with right name of appendix
  "Appendix B: Proofs of Lemmas"

P6-P9
  no changes since LC

P10 ("This means")
  minor words, OK

P11
  "example given in section 1.5" - add a link (& title maybe)

P12-P13
  minor words, OK

P14 ("The interpolation lemma completely ...")

  "check that there is some instance of the entailed graph which is a
  subset of the merge of the original set of graphs."

  Isn't that a set membership check? (ie member<>set rather than
  set<>set, a subset relation)  so shouldn't that be:

  "check that there is some instance of the entailed graph which is a
  member of the merge of the original set of graphs."

P15-P21
  OK - rewordings fine


3. Interpreting the RDF Vocabulary

[Hmm, all new here!]

P1
  OK

P2
  missing: rdf:value

  It's still there, we didn't remove it :)

P3
  "RDF concepts and abstract syntax document[RDF-CONCEPTS]"
  => "_RDF Concepts and Abstract Syntax_ [RDF-CONCEPTS]", link & title

  earlier (1.2 last para) you used "XML data" rather than
  "The corresponding value" or he "XML value of".
  Yous seem to have "XML data values" also in red.

P4
  I think I'd prefer to see "I of (V union crdV)" as in the LC doc,
  since this can be hard to read depending on how the web browser
  breaks the lines.

  Hmm, it isn't clear that these are 3 tables, at least on my
  screen.  You must either alter the formatting or preferable do that
  as well as give them titles.

P5-P7 (3 tables/conditions)
  OK, titles please

P8-P9
  OK. I note that rdf:value does appear here !

P10 "The first condition"
  OK.  This paragraph might be better directly after the tables and
  before the axiomatic triples.

P11
  "section 3.3 below" => link please

P12
  OK

P13
  again, prefer to have the sets on separate lines

P14 (figure 2)
  OK

P15-P15
  OK, minor changes since LC
  
3.1. RDF entailment

P1
  prefer (S union E)
  link section 2 wording

P2
  link lemmas in section 2

P3
  OK

P4
  link to 7.2


3.2. Reification, Containers, Collections and rdf:value

P1-P2
  OK

3.2.1 Reification

P1
  minor change since LC

P2-P9
  unchanged

P10 ("and y is I(aaa)...")
  OK (New last 2 sentences)

P11-P13
  OK, minor word changes

P14-P17
  unchanged


3.2.2 RDF containers

P1-P6
  minor word changes OK

P7-P9 (example)
  unchanged

P10-P11
  minor word changes OK

P12-P16
  unchanged

3.2.3 RDF collections

P1-P3
  minor word changes OK

P4 (example)
  OK - replacement of rdf:type rdf:list with rdf:first

P5
  minor word changes OK

P6-P8
  OK - replacement of rdf:type rdf:list OK.

P9 ("Also, RDF")
  minor word changes OK

P10
  OK - replacement of rdf:type rdf:list OK.

P11 ("It is also possible")
  minor word changes OK

P12
  minor word changes OK

P13
  OK but might be improved by using "any subject of the rdf:first
  property" etc.


3.2.4 rdf:value

P1 "RDF primer" => "_RDF Primer_" with link, title

P2
  ah, disclaimers.  You don't need to write in the third person so
  obliquely  this is meant to be a WG document, so maybe
  change "the editor is unable" to "it is not possible"

  Personally I don't like this dire warning, since rdf:value has been
  in the RDF M&S with a decent use case.  But I guess I lost that one.


4. Interpreting the RDFS Vocabulary
 
P1-3
  minor word changes OK

P4
  changes OK
  "table below" => link it when you make the semantic condition table
  have a title, linke

P5
  changes OK

P6 first table
  unchanged since LC, apart from if and only if :)

P7-P8
  range/domain changes OK (paragarphs seemed to just be swapped
  around from LC version)

P9-P10 
  subPropertyOf changes OK

P11-P12
  subClassOf changes OK

P13
  unchanged

P14
  I think this is OK, datatypes hurt my head.

P15-P16 axiomatic triples
  These seem in rather randon order that is difficult to compare.  I
  assume this is just the old ones + the rdf ones.

P17-P18
  words ok

P19
  I can't confirm these are OK, they look plausible.

P20
  OK


4.1 Extensional Semantic Conditions (Informative)

P1
  OK

P2-P5 (tables)
  I recognise from LC version and match them

P6
  OK

P7
  link to 7.3


4.2 A Note on rdfs:Literal

(was 3.3.1)

P1-8
  OK minor changes from LC


4.3 RDFS Entailment

(was in 4.2)

P1
  OK, link to section 2 wording

P2-P4
  OK


5. Datatyped Interpretations

[substantially different text, diffing impossible]

P1-P8 ("In stating")
  OK

P9
  "RDF concepts and abstract syntax document[RDF-CONCEPTS]"
  => "_RDF Concepts and Abstract Syntax_ [RDF-CONCEPTS]", link & title

  Link XSD part 2 also

  This reads as if XSD are built into these semantics rather than
  being one compatible set of datatypes.  I'd prefer to see something
  along those lines here.  Maybe a new paragraph here and continuing
  with the "The datatype map" sentence.

P10-P11 (XSD datatypes)

  xsd:NMTOKEN is allowed but xsd:NMTOKENS isn't?

  The rest of the list matches XSD datatypes OK.

P12 ("If D is a datatype map")

P13
 OK

P14-P18 (tables)

 [[if <aaa,x> is in D then for any typed literal "sss"^^ddd with I(ddd) = x ,
   ifsss is in the lexical space of x then IL("sss"^^ddd) =
   L2V(x)(sss), otherwise IL("sss"^^ddd) is not in LV
 ]]

 couple of things:
   1) typo: "ifsss" => "if sss"

   2) what is "L2V(x)(sss)" ?  L2V(x) gives you a set (value space)
      so what does "(sss)" do to that set?

P19-P20
  OK

P21 ("The third condition")
 this is the point that you were refering to earlier about
 non-literal values being present in LV and should be linked to from
 there.

 Whereas here you say (on bad xsd-interpretations):
     [[sss ... is not in ICEXT(I(rdfs:Literal))" ]]
 in condition 3 you say (for the general D-interpretation case):
     [[IL("sss"^^ddd) is not in LV]
 are those the same?

P22-P28
  OK

P28
 "Datatype clashes and assertions that ill-typed literals are of type
 rdfs:Literal are the only inconsistencies recognized by this model
 theory."

 This confused me for a while but is correct.  It would IMHO read
 better the other way around:

 "The only inconsistencies recognized by this model theory are
 "datatype clashes and assertions that ill-typed literals are of type
 rdfs:Literal."

P29-P33
  It seems OK but I'm wondering why this is here?  Why do you need to
  talk about larger datatype maps or special graphs where all URIrefs
  are datatypes?  Is it in order to get to the point at the end of
  the last paragraph " adopting this requirement as a formal
  entailment principle would violate the general monotonicity lemma"

  Well, I felt, so what? :)  If you added other requirements, that
  might happen!


5.1 D-entailment

OK


6. Monotonicity of semantic extensions

[wasn't that what you were discussing at the end of section 5?]

OK


7. Entailment rules

OK

7.1 Simple Entailment Rules

se1 & se2 - nnn is a blank node _identifier_ not a blank node
or stick with bnodeID

P3
  "on the same URIref or literal bbb or vvv"

  By your definition, bbb cannot be a literal only a URIref

  maybe:  "on the same URIref bbb or literal vvv" ?

P4-P10
  OK


7.2 RDF Entailment Rules

rdf1 as in LC
rdf2 seems OK

RDF entailment lemma -
  OK, but link to the rdf axiomatic triples by section & link.  Link
  to proof.


7.3 RDFS Entailment Rules

rdfs1 ok

rdfs2, 3, 4a, 4b, 5 (was 5a), 6 (was 5b), 7 (was 6), 8 (was 7a), 9,
rdfs10 (was 7b), rdfs11 (was 8), rdfs12 (was 10)
  in LC doc

rdfs13 ok

RDFS entailment lemma -
  OK, but link to the rdf axiomatic triples by section & link.  Link
  to proof.

rest is OK


7.3.1 Extensional Entailment Rules

Surely this section is Informative since 4.1, on which it is based,
is informative?


P1 add link to section 4.1.

These rules match some of the old ones (pre the weaking) such as
rdfs2a, rdfs3a.

ext1 to ext9 seem plausible but I'm not worrying about them since
they are informative.


7.4 Datatype Entailment Rules
(was 4.3)

P1
  minor changes OK

P2
  changes OK

P3
  spelling - "dataype"

P4 (rdfD1)
  change OK

P5-P10 (rdfD4)
  minor wording (not rules) changes OK

P11-P14
  minor changes OK

P15-P20 (examples)
  minor changes OK

P21
 OK


Appendix A: Translation into Lbase

(status?  informative?)

Not reviewed but looks like a lot of changes.


Appendix B: Proofs of Lemmas

(status?  informative?)

Not reviewed but looks like a lot of changes.


Appendix C: Glossary of Terms 

The minor word changes since LC are fine.


References

[XML-C14N] link broken
should be http://www.w3.org/TR/2002/REC-xml-exc-c14n-20020718/

The style needs a bit of polish.  You need to have 'latest version'
links for the W3C documents.  See the other WDs for examples (to cut
and paste!)
Received on Thursday, 24 July 2003 07:22:15 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:58:47 EDT