notes and comments on RDF syntax and schema documents

Review and comments on:
	RDF Model and Syntax Specification (version July 20, 1998)
		WD-rdf-syntax-19980720
	RDF Schemas (version July 17,1998)
		WD-rdf-schema-19980717

Reviewers:
	Rudi Studer
	Jan Jannink
	Gio Wiederhold
	Erich Neuhold
	Vasan Pichai

Dear People,
     During the course of a meeting with R. V. Guha in which he
presented to us the state of RDF, he urged us to forward our
questions and comments to you the editors of the above documents.
We hope that our input is helpful to you.

We would also like to alert you to a proposal by Deutsch, Fernandez,
Florescu, Levy and Suciu:  A Query Language for XML.  The authors
propose two semantic models for XML which are highly similar to RDF.  
We feel that it is important to ensure that XML-QL is properly
related to RDF, so as to prevent any further fracturing of the XML
community.  Please find the draft at:

http://www.research.att.com/~mff/xml/w3c-note.html


Sincerely, Jan Jannink


----------------------------------------------------------------

WD-rdf-syntax-19980720


2.1 Basic RDF Model

    We appreciate the analogy between an RDF model and an ER diagram.
However, RDF models also include instance level information, whereas
ER diagrams are purely at the schema level.  It would help to
contrast the two in the text.  This is a good place to introduce
the RDF schema effort.

    The equivalence relation between two RDF expressions is mentioned
here, and certainly deserves also a section in the formal model.

    The term 'statement' is introduced in the definition of the three
fundamental object types of RDF.  Elsewhere, the term property is used.
We find the terms 'property type' and 'property' are often swapped
in the documents, and are therefore unfortunate choices.  It is important
that a uniform vocabulary be used throughout the documents, and also
in the BNF definitions.  Here is a list of some terms that are used
informally and interchangeably that cause confusion in the documents.

Terms relating to Resource:
	Object
	Class
	Node
	Element
	Description

Terms relating to Property Types:
	Attribute
	Relation
	Property

Terms relating to Statements:
	Property
	Fact
	Triple


2.2 Basic RDF Syntax

    In the first paragraph metadata and schema are introduced without
a definition in the RDF context.  It would help to direct readers
immediately to 2.2.3, which covers the topic.

    It would help to describe briefly how the two RDF syntaxes differ
in the introductory paragraph, and why they both exist.


2.2.1 Basic Serialization Syntax

    In the BNF grammar, it would help greatly to align the terms
'description' and 'property' to 'resource' and 'property type'
respectively.

    It is important to discuss why embedded attributes like 'ID',
'about', appear in the syntax, while others like 'instanceOf' do
not.  The absence of 'instanceOf', and other special property types
from the grammar seems arbitrary.


2.2.3 Schemas and Namespaces

    We feel that the schemas document should be merged with the
syntax document.  At the least it is necessary to describe in the
Schema document all of the embedded attributes in the Syntax document,
and to use a consistent informal language to describe the concepts
in both documents.  A glossary in the the Syntax document with a
complete list of the special Resources and Property Types, can serve
as the basis for the RDF schema describing RDF (this important item
is not currently present in either document).


4. Statements about Statements

    The reification section is a step in the direction of defining the
RDF schema describing RDF.  The clarity of exposition of reification of
Description elements must be improved.  This section brings up a very
important issue.  If a description is reified as a Bag, then we should
consider Bags to be the fundamental type of resource.  In effect we are
stating that a single resource is a special case (a constrained) bag.
This brings additional context to the discussion in 3.4. on containers
vs. repeated properties.  Also, when bags are the fundamental resource,
then the need for the attribute bagID is reduced.


6. Formal Grammar for RDF

    A formal, systematic presentation of embedded attributes is important.
It is not clear why ID, about, aboutEach, and bagID are different from
other property types.  This topic deserves a subsection of its own.


----------------------------------------------------------------

WD-rdf-schema-19980717


The syntax document points to XML namespaces to specify RDF schemas.
The schema document does not.  Please synchronize on this point.

The document names a Schema Specification Language.  Is it a language
or is it more properly termed a vocabulary?


2. Classes and Property Types

The set of terms that define RDF should not be split into the RDF and
RDFS schemas.  Also, include the collection terms from RDF.


3. Constraints

List the special embedded attributes ID, about, aboutEach, bagID.

3.1.3. RDFS:domain

Is it possible for more than one class to be the domain of a property type?


4. Figures (change section number from 3)


The term 'necessity' is different from usage elsewhere.  Use 'cardinality'
if this term is to remain in RDF Schema.  Also, cardinality should
properly be considered a constraint.

The ordinals should be specialized constraints as well.

Should strings not be considered a resource?  Excluding them in the figure
is potentially misleading.

Received on Friday, 28 August 1998 04:58:39 UTC