Re: Requirements Document

Hi:

I have a bunch of comments on the Web Ontology Requirements document (Feb 7).

There are a lot of places where the wording needs to be changed.  I
have carefully gone over the first section and have noted the worst
problems in other sections, but there are lots more places where the
wording is not correct.

The document suffers from a serious case of self-importance.  In
particular, 

	The Web Ontology language will be a significant advance in Web
	functionality and take interoperability beyond the present stage.

pegged my hype-meter.  I think that we should be very careful to not make
such statements, particularly in a requirements document.

The document still reads as if it was the separate creation of
different people.  The style should be at least a bit more uniform between
the various use cases subsections.

The order of the design goals is rather strange.  In particular, I
would think that it would be better to move 3.4 before 3.3

The document uses words with technical meaning in places where the
technical meaning may not be what is wanted.  In particular,  unless
``resource'' means what it means in RDF I think that it should be avoided.
So
	Ontologies must be resources with their own unique identifiers.
can easily be read as implying that ontologies must be objects just like
regular objects.  

The document anticipates some technical features of OWL.  In particular, it
uses URI as the term identification mechanism.  This brings up the URI vs
QName discussion.  



There are a number of places where the document places very strong
requirements on OWL.  In particular it states
	The language *must* allow properties to be associated with
	statements.
I don't see this requirement on our ``A'' list.  
There are other very strong requirements in the document, including the
closed-world requirement.  This placing of requirements on OWL continues
even into the Objectives section, where it says
	At a minimum, the language should recommend to users how they can
	specify their own default mechanisms.

The beginning of Section 5 goes even further, and calls the requirements in
Section 4 ``minimal [...] features''.  This seems to be setting the WG up
to fail big-time if even one of the requirements in Section 4 is not
totally solved by OWL. 

The lexical representations requirement seems to be arguing that URIs have
to have multiple lexical forms.  I don't think that this is what was wanted
here.


Many of the requirements and objectives are not very well specified.  In
particular, what does it mean to
	support the use of variables in ontology definitions. 

Some of the objectives go beyond a web ontology language.  For example,
I view speech acts and conditions outside of a web ontology language if
there is any semantics attached to their constructs.


Here are some specific wording changes:

It shall provide ...
	The language shall provide ...

for such a language.
	to support these g & r.
	[use cases are not really *for* a language]

i.e. the objectives
	i.e., the objectives

and which tasks support it.
	and which tasks require it.
	[tasks don't support design goals---maybe you could change to
	support the inclusion of]

of standard vocabularly terms
	of standard terms
	[it might be better to keep away from anything with a natural
	language tone to it]

This notion of ontologies comes from Artificial Intelligence, 
	[I don't think that this is really true, and in any case, I don't
	what to ghettoize this view of ontologies]

terms in an ontology are not just defined in some natural language,
they also have logical definitions
	[Again, I would stay away from NL.  Also, the terms in a WebOnt
	ontology won't be defined in NL.]

and attributes defined by DTDs or XML Schemas do not have any semantics
associated with them;  
	[This is a very strong statement.  It might be fine for an opinion
	piece, but do we really want to get all the XML people even madder
	with us?]

In RDF Schema, you can
	RDF Schema can
	[I would stay away from the use of second-person.]

and you can define properties
	and can define properties

richer semantics are needed.
	[statements like this need some justification]

One of the goals of this document is to specify exactly what is
needed by a Web Ontology language
	[I thought that this was an initial list, not the final one.]

There are numerous uses for ontologies on the Web. 
	[This should, if included, be in the first section.]

be the
	be that

large corporations massive 
	large corporation's
	[is massive really needed here]

Language neutral representation
	Language-neutral representation

changing any axioms, thus determining backwards-compatibility requires 
more than a simple comparison of axioms.
	[This assumes that an ontology will contain axioms.]

low-learning barrier
	low learning barrier

an reused
	and reused

Received on Thursday, 7 February 2002 14:47:48 UTC