[
Lists Home |
Date Index |
Thread Index
]
> I agree with Tim that XML is a name/label/structure
> system and as such, doesn't care much about this
> debate. However, that simply says the developer
> has to care, so we still have to face up to the
> symbol grounding problem elaborated in detail
> by Charles Peirce in his papers on semiotics over
> a hundred years ago and clarified in the works
> of John Sowa. Harnad [3] explains it satisfactorily
> in terms of AI approaches including combining
> connection systems (eg, neural netws) with symbol
> systems. All good background, but there are other
> approaches and we should explore these.
Why do we have to face up to the symbol grounding problem? If I
systematically replace "meaningful" with "valid" I can come up with
solutions for namespace composability that are purely syntactic. E.g., James
Clark's NRL.
Why not leave meaning the province of humans, who sometimes write programs
to give an operational "meaning" to XML documents? The meaning is not
intrinsic to the document; only the syntax is.
> In short, clearly namespaces enable composability
> at the syntactic level. Just as clearly, many
> combinations are meaningless.
If you say many combinations are invalid and will not be accepted by some
program, we have grounds for agreement. But if you want to assert that
combinations are meaningful that will not be accepted by any program, I
wonder what is the point?
Truly puzzled but willing to learn.
Bob Foster
> As Harnad says
> when defining systematicity:
>
> "The patterns of interconnections do not decompose, combine
> and recombine according to a formal syntax that can be given
> a systematic semantic interpretation."
>
> So in effect, we can create namespace aggregates
> which are not systematic. So via namespaces,
> any set of XML application productions (by which
> I mean, a production from HTML, from SVG, from
> X3D, or XSLT) can be combined and be syntactically
> correct.
>
> How can one determine:
>
> 1. If a given combination is meaningful
> 2. How to discover that meaning
> 3. How to assign that combination or even a single
> production to a running piece of code
>
> Item three is where the rubber meets the road.
>
> a. Does RDF address these questions?
> b. Is it better for worse particulary for item 3
> than say using stylesheet assignments
> c. Are other approaches such as abstract
> object models as good or better than RDF for
> writing the rules of a semantically valid
> combination?
>
> Next, is it desirable or workable that any
> arbitrary combination of XML productions from
> any language be meaningful? I think the answer
> here is no and leads back to 1.
>
> I think this an important topic because it touches
> on issues such as when should two application language
> working groups seek convergence, can we create
> XML application languages that don't set of IP tripwires
> by ensuring implementations based on IP aren't a part
> of the language definition, should we begin to classify
> semantically valid XML production combinations, and where
> in that will standardization impede innovation,
> is it really a good idea to use a standard namespace
> name to point to running code?
>
> len
>
> [1] http://tbray.org/ongoing/When/200x/2003/08/11/SymbolGrounding
> [2] http://weblog.infoworld.com/udell/2003/08/11.html#a775
> [3]
http://www.ecs.soton.ac.uk/~harnad/Papers/Harnad/harnad90.sgproblem.html
|
