 | |
RDF untyped
literals: tidy? | | |
 Info designer
issues
 Dublin Core was underspecified
almost to the point of uselessness
Suggests an untidy literal view of the world: information given carries minimal meaning until
further qualified. In some cases, further qualification may be local
assumption (leads to non-monotinicity when formalized).
Different rules for different names in the
language (URIs and untyped literals)
PatH condensation of my point.
Reinterpretation of properties
Consider:
Jenny age "10" .
Brian suggests defining age property such that:
-
if its object is a literal then it denotes the numeral representing the subject's age in years
-
if it is not a literal then it denotes the number that is the subject's age in years.
Thus, the relational extension of age contains both <Jenny,"10"> and <Jenny,10>
This is fine, but what about string-valued properties. Then a relational extension defined in
this way would give some unintended truths.
Suppose a datatype is enclosed-string, an occurrence of which denote the sequence of characters contained
in containing '<<' and '>>'. Assert:
Jenny name "<<Jenny>>" .
also, I *define* a URI <http://mydomain.org/string/Jenny> to denote the string "Jenny". Then say:
Jenny name <http://mydomain.org/string/Jenny> .
If both of these statements are true, what is contained in the relational extension of 'name':
<Jenny,"Jenny">
<Jenny,"<<Jenny>>">
I can then conclude the truth of:
Jenny name "<<<<Jenny>>>>" .
| 
