on RDF concepts: thumbs up, contingent on a few things

Bottom line: thumbs up, contingent on ~5
small but CRITICAL problems (for each of these,
I propose a fix that I expect is uncontroversial*);
and contingent on assurance from other parties
that the detailed technical bits after section
2.3.2 are OK.

(*hmm... actually, looking over these again,
the "RDF doesn't depend
normatively on RDFS" comment below,
and its fix, could be controversial.)

To elaborate: I read and very carefully reviewed
and commented on the introductory and motivational
material, up thru 2.3.2 Social meaning.
I have read the rest of section 2.3 Meaning of RDF documents,
with less attention to detail a few earlier
times, and I'm satisfied that it's valuable
stuff, if not finished. I specifically invited
review of 2.3 from a few of my colleagues (you've
likely seen the comments from Sandro by now,
as well as TimBL's earlier review,
and I got Danny to look at the use of
legal terminology; he had no objection; he
was considering some suggestion about the
liar's paradox... one that I didn't think
was critical, and that he agreed to send
mail to www-rdf-comments if he thought it
was really critical).

I expect other reviewers are closely poring over
the technical details of the abstract syntax and such.
I'm sorry, but I have run out of attention to detail
at that level; I cannot provide assurance to the WG that
it's (a) consistent with WG decisions to date
nor (b) just plain correct in detail.


That's it for bottom line. The remainder
of this message is paper trail and detailed comments.


*** Papertrail:

what I reviewed:
http://lists.w3.org/Archives/Public/www-archive/2002Oct/att-0066/02-rdf-concepts.html

how I found it:

# RDF Concepts and Abstract Data Model - Review Copy
Jeremy Carroll (Sat, Oct 26 2002)
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Oct/0389.html

(by the way: it wasn't all that easy to find;
I thought I could just look at anybody else's
review comments and follow their pointers to
the current draft; but nobody else seems to
cite exactly what they reviewed.)

wherein I agreed to review it:

ACTION: PatH, DanC and JosD to review the updated version of the
concepts
document
  recorded in http://www.w3.org/2002/10/25-rdfcore-irc#T15-32-42
# Minutes of RDFCore WG Telecon 2002-10-25
Jos De_Roo (Fri, Oct 25 2002) 
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Oct/0386.html


*** Comments as I read it:

CRITICAL denotes must-fix-before-release issues.
I give suggested text in each case.

WRONG denotes things that I think are factually/objectively
wrong, though I don't propose to put them in
the critical path for publication

the rest are just editorial suggestions.

I hope the level of detail in my comments communicates
how valuable I think this document is, rather than
suggesting that I think it's somehow broken.


... Abstract <WRONG>

"The Resource Description Framework (RDF) is a data format "

er... no, the framework is a framework.

(occurs again in intro)

... </WRONG> <WRONG>

"Web resources, and other information"

er... resource is the minimally constraining
class; there's nothing outside it. This is
incoherent.

... </WRONG>

"This document defines the abstract graph syntax"

which abstract graph syntax? i.e. to what
does that definite description refer?
suggest:
	... an abstract syntax ...

...

ooh; I like the next bit:

"which serves to link its XML serialization to its formal semantics"


...

I suggest this can and should be phrased positively:

"It also describes some other technical aspects of RDF that do not fall
under the topics of formal semantics, XML serialization syntax or RDF
schema and vocabulary definitions (which are each covered by a separate
document in this series)."

suggest:

  This document presents the overall framework;
  it is supplemented by detailed specifications
  of serialization in XML and formal semantics,
  and by a primer and a test collection.

... <CRITICAL>

"The normative documentation of RDF falls broadly into the following
areas:
 ...
 # RDF vocabulary definition language (RDF schema)"

no, the normative documentation of RDF doesn't
include RDFS. Strike that bullet, or move
the mention of RDFS ouside the list, ala:

	The framework is designed so that vocabularies
	can be layered on top of this core. RDFS is the first such
	vocabulary. Others (cf OWL and the applications
	in the primer) are in development.

(This point suggests that the RDFS part of the model
theory should be split out from the rest on process
grounds. But I'm not reviewing that document here...)

... </CRITICAL>

ambiguous pronoun reference:

"RDF is a member of the family of languages that use XML, which in turn
provides a syntactic framework for representing documents and other
information. It has a simple graph-based data model"

Does 'it' refer to XML or RDF?

...

"well founded deductions in RDF data"

is 'well founded' used here as a particular term of art?
I don't see what it adds.

...

"The real value of RDF comes..."

Good point, but I don't like the phrasing.
My brother told me, when he was in high school journalism
(he's 7 years older than I am, so this was quite a while ago)
that "very" is almost always superfluous. I think "real" here
is like that. suggest:

	RDF is designed to represent information in
	a minimally constraining, maximally flexible
	way. While application-specific XML
	formats are often more perspicuous,
	using a general purpose framework facilitates
	sharing data between applications. The value of
	information thus increases as it becomes accessible to
	more and more applications across the entire Internet.


...

"2.2.1 A simple data model

RDF has a simple data model that is easy for applications to process and
manipulate."

Hm... if it's that simple, can't we say *a little bit*
about what it is in this section? Maybe...

	RDF has a simple data model, similar to
	an directed-labelled-graph, or 3-column
	relation, that is easy for applications...

and I might substitute "straightforward" for easy.

<diversion>

I wonder if "relational" is better than "graph" when
talking about the abstract syntax or datamodel or
whatever it is.

Thinking about RDF as the simplest SQL table that can
subsume all other SQL tables might be a useful
rhetorical device.

</diversion>

...

change
  logicians call "model theory"
to
  "model theory" as used in logic literature.

...

"a sound basis for reasoning"

er... "sound" is a term of art, like "model theory".
strike it?

This principle 2.2.2 about formal semantics is
insufficiently motivated. From discussion with
TimBL, I gather there's a scalability benefit
to using rigorous formal semantics: for at
least this existential-conjunctive fragment
of FOL, we're sufficiently confident with it
that we trust a derivation of 1000 steps
just as much as we trust a derivation of 1
or 2 steps. We understand it well enough
to delegate to machines based on it.
Something like that.


... <CRITICAL>

"This is where RDF departs from the XML approach to data
representation, which is generally quite prescriptive"

No, if there is one XML approach to data representation,
it by definition includes RDF/XML. Reprhase as:

	This is where RDF departs from more prescriptive
	approaches to representing data in XML ...


... </CRITICAL> <WRONG>

"But what consitutes a "simple fact"? Roughly, the kind of
information that can be stored in a relational database"

suggest:
	can be stored in +one cell of+ a relational database

... </WRONG>

"Relationships involving more than two things can be expressed as a
conjunction of binary relations"

ought to stipulate that this is awkward, ala

	Relationships involving more than two things
	are awkward, thought straightforward, to
	express as a conjuction ...

...

"... but the use allowed by RDF has first-order semantics."

cite the relevant MT section.

... <CRITICAL>

"... RDF aims to provide for universal expression ..."

don't go there. Try "flexible expression" or some such.

... </CRITICAL>

ah; perhaps 2.3.1 provides the motivation I was looking
for under 2.2.2. forward reference, please. Hmm...
no, it doesn't include the scaling motivation after all.

...

"the more tightly reality is circumscribed"

this use of 'circumscription' might be confused with the
term of art introduced by McCarthy.
suggest: constrained.

...

"no such assertion is considered to be made"

s/considered to be//. strunk-n-white.

... <CRITICAL>

"Thus, there is a distinction between RDF expressions that are asserted,
and those that are not."

That looks like a good defining occurence of
'asserted'. Please make it a hyperlink target
and mark it up with <dfn>
(and point to it from a glossary at the end?).

... </CRITICAL>

s/is being registered for indicating/indicates/.

...

I suspect an example is necessary to make this point
sufficiently clear:

"This means a graph may contain "defining information" that is opaque to
logical reasoners. This information may be used by human interpreters of
RDF information, or programmers writing software to perform specialized
forms of deduction in the Semantic Web."

... <CRITICAL>

"If you publish a graph G and G logically entails G'"

There's more than one entailment relationship in
this framework. I'm not sure how to be more clear
here, but I know it's important for layering
issues, esp. WebOnt coordination.

Please put one of those [[[we know this needs work]]]
markers there.

... </CRITICAL>

"Many inferences are performed by processes, embedded in software
implementations, whose validity is not formally demonstrable, and must
be assumed or trusted to be socially acceptable."

pls mention another option, in addition to assumed
or trusted:

	... or assured, by way of digital signature, ...

...

I exhausted by review-eye-balls at this point,
and sorta skimmed the rest.





-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Friday, 1 November 2002 00:49:14 UTC