Re: [xml-dev] more on "subelement significance"

[Mike Moran <mmoran@netphysic.com>]

Bob Hutchison wrote:

> On 01/10/03 4:20 PM, "Seairth Jacobs" <seairth@seairth.com> wrote:
> 
>> The usefulness of any given subelement is due to the knowledge of its
>>namespace, document type, and/or parent element.  Without any of the three,
>>the subelement does not have a useful meaning.
>>
>>
>>Thoughts?
>>
> 
> I'm with Simon on the convulsions. Here are a few thoughts that jiggled
> loose. I'm not sure I've come to my senses yet, they seem a bit cluttered,
> but here goes anyway :-)
> 
> XML is a data representation. What's the meaning of data? Does data contain
> its own meaning?
[ ... ]
> 
> What if I showed you this XML document:
> 
>     <root>
>         <things>
>             <thing name="one"/>
>             <thing name="two"/>
>         </things>
>         <people>
>             <person name="Jack">
>                 <owns thing="one"/>
>             </person>
>             <person name="Jill">
>                 <owns thing="two"/>
>             </person>
>         </people>
>     </root>
> 
> Where did the 'meaning' of <thing> happen for you? Did <things>, <root>, or
> <people> contribute anything to your understanding? Did the use of English
> words contribute anything? Consider:
>     <a>
>         <b>
>             <d e='one'/>
>             <d e='two'/>
>         </b>
>         <c>
>             <h i='j'>
>                 <l m='one'/>
>             </h>
>             <h i='k'>
>                 <l m='two'/>
>             </h>
>         </c>
>     </a>
> 
> This has the identical 'meaning' in XML as the previous example. It's just
> that the human interpreter cannot interpret it very well.
[ ... ]


Isn't this sort of thing why RDF exists at all? The association of 
semantics with data is done by a grounding of some terms in `real' 
objects and others in rules for working on those objects. Semantics 
arises through action and definition. In principle, I suppose you could 
take a set of facts, A, B and C and a set of rules A => D, D & C => F, 
and then find a set of real world inferences which this rule chain is 
isomorphic to, but the particular chain you see it as depends on what 
names you assign to the rules and the facts. There is no way round this.

This is exactly the same problem as above ie two isomorphic structures 
with a different `meaning'. RDF is just a regularised and standardised 
way of apply names to rules and facts. You have to accept that when you 
see <person> or uri:person, this actually corresponds to the entity in 
your head which *you* see a person as.

If you have a standard set of rules or inferences then you can 
`validate' that the person on the page is the same as the person in your 
head by saying things like `well, does this <person> have a <name>?". If 
<person> actually has <wheels> then you may be less likely to say this 
is a real person. However, this is not a gauranteed route to certainty 
of isomorphism since, for example, this person may be in a <wheelchair>.

As Simon pointed out, it is all very much in the eye of the beholder, 
but you get quite far by having a standardised eye ie a set of agreed 
upon inference rules and names.

</waffle>

--
Mike