Lists Home |
Date Index |
Jeff Greif wrote:
> I was attempting to make a pragmatic point, not a criticism of the RDDL
> language. If I deference a namespace URI and find nothing, I don't know
> whether it was intended by the doc author that there be nothing there, or
> there was supposed to be something there but the URI had a typo. If the
> host mentioned in that namespace URI is down and I'm hoping to find, say,
> collection of stylesheet resources, or a jar file to use to process the
> original document, I don't know whether to try again later or look
> else. If I never dereference the namespace URI, I won't know what useful
> things I might have missed. That's why I'd like to know whether to expect
> to find something or not, perhaps by some additional attributes, by some
> convention involving the URI type, or something else.
Fair enough. The convention _should be_ (IMHO) that URIs that have schemes
associated with network protocols are resolvable using that protocol. So
that best practice is to associate all "http:" URIs with some document for
the reasons you describe above.
Similarly use of a "urn:" URI might indicate that you are unlikely to find a
document on resolving the "urn:" scheme based URI. On the other hand there
are people who are quite interested in the development and promotion of
'URN's and rightly claim that with the proper software one can resolve a
"urn:" based URI. Similarly there are no police that will arrest you if you
fail to place a document at the end of an "http:" based URI -- though Dan
Connolly has been known to chastise people on some occasions.
Then there are arguments about what one should find at the namespace URI.
Ought it be a Microsoft Schema (I forget the name of its early XML schema
language -- but I recall it was somehow attached to the namespace URI at one
point)? Ought it be an XML Schema? Ought it be an RDF Schema? etc.
So I say: yes you are correct, "http" URIs _ought_ to dereference to a
document by convention, and RDDL is a solution to what such URIs ought
You preferred the convention be that _nothing ever_ be returned. I prefer
the convention be that _something always_ be returned.