Re: (tentative) container model proposal

On Wed, 6 Jun 2001, Dan Connolly wrote:

> Dan Brickley wrote:
> >         http://www.w3.org/2001/05/rdf-c/ (copied below in text form)
> [...]
> > I am well aware that the proposal is currently in a skeletal state, and
> > lacks test cases. I hope it provides at least some useful background for
> > discussion on friday. Regarding test cases, I have to say I'm stumped: the
> > implementations this relate to are databases, APIs etc rather than
> > parsers, and I am not sure how best to represent tests of that kind.
>
> Er... maybe I'm missing something, but I read it twice, and it
> seems to reduce to this simple test case: is this an RDF
> document or not?

I have heard it suggested (on www-rdf-interest and personal communication)
that, whilst this sort of file is an RDF document, the language of
paragraphs P189 thru P193 of M+S (see
http://lists.w3.org/Archives/Public/www-archive/2001Jun/att-0021/00-part)
imposes behavioural constraints on RDF databases, query engines etc. For
example, I think there was some discussion regarding the Mozilla RDF
API's "proper" behaviour if a element is removed from a list via the API.

Some people have appeared to say "yes this is an RDF doc" to your test,
yet believe P189-P193 impose behavioural requirements on RDF
implementations.

>
> <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
>     <rdf:Description>
>         <rdf:_1>abc</rdf:_1>
>         <rdf:_3>def</rdf:_3>
>     </rdf:Description>
> </rdf:RDF>
>
> The simplest answer that meets my needs is: yes, that's
> an RDF document. The parts of the spec that suggest
> otherwise are an error.

I agree. Would you support the removal of P189-193 from the "RDF formal
model" section of the spec?

> To my mind, anything that suggest that containers are
> fundamental in any way -- that they are anything
> more than a standardized vocabulary of classes and
> properties and some syntactic sugar -- is an error.

Indeed.

> Now this doesn't resolve either of the active issues you
> own, danbri, so maybe I missed the gist of your proposal.
>
> http://www.w3.org/2000/03/rdf-tracking/#rdf-containers-syntax-ambiguity
> http://www.w3.org/2000/03/rdf-tracking/#rdf-containers-syntax-vs-schema

[ * danbri does some ACTION archaelogy ]

This appears to have mutated slightly since I was originally ACTION'd in
the first meeting; my apologies for not clarifying things sooner.

http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001May/att-0000/01-2001-04-27.html

[[
	Issue Discussion
	We discussed 3 technical issues:

	#rdf-ns-prefix-confusion
	The XML Namespace spec doesn't address unprefixed attributes. We'll have
	to live with their bug. RESOLVED: We'll strongly recommend the use of
	namespace qualified attributes, and allow but deprecate unqualified
	attributes.
	#rdf-container-syntax-ambiguity
	#rdf-container-syntax-vs-schema
	Ora noted that it was an original requirement for RDF that parsing should
	not require knowledge of the schema.
	The discussion on the latter 2 issues largely occurred in parallel and
	became a more general discussion about containers. Several people
	advocated deprecating container types and/or making them a separate module
	(this also applies to reification).

	It also brought out several larger issues:

	RDF Syntax vs. Model
	We should record more of the history behind various RDF issues.
	What is the real-world use of containers and other RDF features? What will
	break if we change stuff? ACTION (everyone): send RDF feature usage
	information from real-world applications you know of to Guha, who will
	collect it.
	Modularization can be accomplished via namespaces or documents.
	Redefinition of the RDF(S) namespaces may not be required.
	ACTION: Dan Brickley will own the "container problem". Ora Lassila will
	help.
]]


My action (followed up after the call in a jetlagged discussion with
Ora in Hong Kong) was a follow on from the minuted observation that "The
discussion on the latter 2 issues largely occurred in parallel and became
a more general discussion about containers". The "container problem" that
I went away with Ora to summarise was to look at some of the
broader-than-syntax problems with containers that we touched on in that
call, and come back with an account of the model-oriented (rather than
pure syntax) problems. We focussed on the privileged role given to
containers in the model.

Having convinced myself there is indeed an issue here, I believe it
deserves a separate place in the issue list.

Proposed issue summary:

rdf-containers-formalmodel
[
Parags 189-193 of M+S suggest a privileged role for RDF containers within
the formal model at the heart of RDF. Furthermore, they suggest
largely unimplemented (**need to hear about Jan's
implementation**) constraints, either on XML encodings of RDF, on other
(eg. database implementations) or on both. These paragraphs are either in
error (RDF does allow for partial descriptions) or editorially redundant:
in either case they should be removed.
]

Dan

Received on Thursday, 7 June 2001 18:33:06 UTC