[
Lists Home |
Date Index |
Thread Index
]
Hi Len,
Even if this look navel discussion, in the context of what's happening
with XML and the level of dissatisfaction we feel with the current
direction. Maybe this kind of discussion may help us go back to the
basics and what is good about the Web. Simply that: the web, the
capacity to link things. This is why the linkage issue is so important.
Especially in this decade in which we will have to link multimedia stuff
with textual information (at least when broadband will take off -
Anyway, Asia and the competitive pressure will help us to move toward
that goal)
Didier said:
>Perhaps a link is at its more abstract level simply indicates a
>relationship and when mapped to an element,
Len replied:
It can, but a link like that is a named set of properties.
Didier replies:
As in the Grove model. A grove entity is a property set. So in the
abstract, a rendering object is a grove entity as well as is a semantic
element such as a bank account. Both have a property set attached to
them. Said like that, a link a simply a linkage between two property
sets. Sounds like the basic goal of the semantic web. What is missing
though, is to bring to explicit knowledge the property set of these
entities. Schema language or DTDs can do that for marked documents but
for a video or a song we have no ways to explicitly state the internal
structure either encapsulated or not with any agent. Or the question is:
Is a schema sufficient to define these property set and the relationship
they have? At least a good question to pose and probably a good thesis
theme.
What is good about a model based on objects and property set is that we
do not have to use a finite set of types (like boolean, numbers,
etc...). The objects are defined by the property set. In some ways a
property set tells more about the object than a simple axiom like saying
this is a number. Saying that this is a number says very little about
the nature of the number but making explicit its properties can say a
lot more about it. This is a phenomenological approach (oops sorry for
philosophy allergic fellows) since we discover the object through its
facets. And we process or manipulate this object with a certain
intention, this implies that some properties may be more important than
other dependent on the context.
This said, I think that W3C (and its not the fault of its founder or its
member), by nature and diverse communities of minds it represent, is a
strong candidate for schizophrenic behavior. On the one hand, it
publicize the semantic web model based on objects and property sets and
on the other hand do not use this rich model for its own purpose and
model definition. For instance, having an SVG object model presented as
property sets and relationship between these objects. We have a text
document encapsulating the knowledge but no formal document consumable
by a processor.
I think that some of us in the XML community made already the inference
that a type could be described by an RDF description. I personally would
try this path in Didier's labs to see if this is useful even if my first
try is criticized. But hey, we learn by taking risks.
So, in conclusion, I think that what the Grove model gave us is that a
type or an object can be described by a property set and that property
set can be linked together. See, a single sentence that could express
something as the semantic web as well as other views of the web.
I wont' play whitehead (or maybe try to do :0 ) but will try an info set
definition:
A document is the basic unit of the web. In the abstract, a document can
be envisioned as a collection of objects and each object more
particularly defined by a property set (i.e. a collection of
properties). Thus, a particular object instance can be thought as a
property set associated with specific values. A link is a reference to
another object instance either in the same document or in one or more
external documents.
Issues:
- Can Xpointer satisfy such reference to objects' instance linkage?
(especially in the case of linking movie or song fragments)
- Can RDF be a good language to define the property sets and in some
ways be a remake of grove plans since it is fashion to redo stuff every
decade.
- Can we go over strict types and be able to consider stuff like
circles, bank account, numbers etc.. as types. Said differently can we,
as a community, go beyond the simple types expressed in procedural
languages and relational databases?
- Can the cobbler be well equipped with shoes and start walking the talk
by recommending to start the semantic web with its own stuff? You know
who the cobbler is don't' you?
- other issues I do not think for the moment but that my dear fellows
have in their minds.
Sorry for the navel discussion :0
Cheers
Didier PH Martin
|
