xml-dev - Re: SV: A certain difficulty - lack of action!

[Sergey Melnik <melnik@db.stanford.edu>]

David Megginson wrote:
> 
> Greg FitzPatrick writes:
> 
>  > So like .. er, we shouldn't do anything at all?
> 
> Well, sort-of.  Let's wait a while and see what the implementors come
> up with, and then standardize *after* we have real-world experience
> rather than trying to do it in advance.
> 
> If the implementors don't come up with anything significant, then it's
> probably better to throw out RDF altogether and start over.  I'm still
> tempted to design something RDF-like but much simpler and better
> documented [...]

With due respect, I think that it is hardly possible to design a data
model that is simpler than the one of RDF. Your next email [1] only
confirms that fact. The two most important contributions of RDF are
object identity and relationships. The need for them in the XML world is
so apparent that I believe these features will show up in some
XML-related standard very soon. I agree with you that the RDF data model
itself is *slightly* overloaded. Thus, Bag, Alt and Seq do not seem to
belong to the core of the model. However, I would argue that rdf:type,
reification and ordinal properties (_1, _2, ...) are absolutely
essential for the representational completeness. Whatever RDF-like
standard will be widely accepted, those features will be needed at some
point in the same way as identity and relationships are emerging in XML.

My answer to the what-to-do question:

It has been widely recognized that the current RDF syntax is buggy and
confusing. Since it is part of the spec, it makes the whole spec bogus.
On the other hand, no significant complaints have been uttered with
respect to the data model. Therefore, the first low-effort step that I
would like to see is an RDF spec amendment that strips out all
serialization from the spec. It would reduce the spec to a dozen of
easy-to-understand pages. No special "group of experts" is required for
that, just a couple of editors. This amendment would not introduce any
changes to the model.

As a next step I'd like to have a "poorest man" RDF/XML serialization:

<rdf>
  <triple subject="[URI]" predicate="[URI]" object="[URI]"/>
  <triple subject="[URI]" predicate="[URI]">[PDCATA]</triple>
  ...
</rdf>

This serialization is unambiguous, self-explaining, and can be
implemented using 10 lines of code. It would provide a good reference
point for determining what features in the current RDF/XML serialization
are syntactic sugar (e.g. aboutEach) or bugs (i.e. not reflected in the
model like xml:lang, aboutEachPrefix etc.)

After having an undisputable basis for interoperability one could think
of more "human-oriented" serializations. However, any given serialized
model S would be required to pass the correctness test:

M1 = parse(S)
M2 = parse(serialize(M1))
M1 == M2 ?

I highly appreciate the cognitive effort that W3C invested in RDF and
feel bitter about it being widely ignored due to the lack of further
action of the consortium. It is sad that for this reason folks like you,
David, have to consider reinventing the wheel.

Sergey Melnik

[1]
http://lists.w3.org/Archives/Public/www-rdf-interest/2000Feb/0114.html

***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/threads.html
***************************************************************************