[
Lists Home |
Date Index |
Thread Index
]
Hi John,
John said:
That is intentional in the RDF Schema design: properties are a primitive
notion distinct from classes. This allows you to add your own properties
to existing classes. For example, you might define the class your:book,
with properties dc:title, dc:author, isbn:isbn, and your:rating, the last
representing what you think of the book. I could then add new properties
to the class such as my:rating or my:criticizesW3C. This is part of the
support machinery for the design principle "anyone can say anything about
anything in RDF".
--
Didier replies:
I can also bring on the table that an RDF instance expressed as:
<rdf:description about="#myUniqueIdentifier">
<property1>value1</property1>
<property2>value2</property2>
<property3>value3</property3>
</rdf:description>
This statement is located in one file. Let's then have another RDF statement
in another file:
<rdf:description about="#myUniqueIdentifier">
<property4>value1</property4>
</rdf:description>
I don't include here scenarios where a specific property value in one file
is different form the value in the other but more the case, when a new
property and its associated value is specified in another file. RDF allows
me to add this new "triple" to the resource and come out with the following
result
<rdf:description about="#myUniqueIdentifier">
<property1>value1</property1>
<property2>value2</property2>
<property3>value3</property3>
<property4>value1</property4>
</rdf:description>
Now, I let you make the right inference. See, it is possible to have
properties "attached" to an "instance" and still have a distributed
definition. We can apply the same reasoning to classes and properties
definitions. I know, editors like protégé consider properties (i.e. slots)
as separate entities from classes (i.e. classes). This way of thinking
percolated in RDFS (but not in RDF since this latter was based on previous
languages like KIF and MCF). Object oriented languages have the advantage of
real life test (i.e. programmers using it) and from usability studies (a la
Ben schnederman). This is why class definitions are compact. Now translating
the composition rule inherent to frames (a la RDF) into classes and property
definitions and integrating what we learned with object oriented notations,
we end up with something more readable and usable. Therefore the question:
is it more convenient or more readable? The answer is no.
PS: Off course, in last resort, we can always use the argument that an
editor will hide that complexity. I would answer that it is needed since the
actual language is not very usable (from the usability perspective).
Cheers
Didier
|
