On 21 Aug 2007, at 17:11, Eric Jain wrote:
[snip]
Here are two typical applications that I know people have been
playing with (and that don't work quite as well if URI's are non-
HTTP or unresolvable):
1. Piggy Bank <http://simile.mit.edu/wiki/Piggy_Bank>, a semantic
web browser and data collection tool.
2. Swoogle <http://swoogle.umbc.edu/>, a semantic web crawler.
[snip]
I thought I'd test this out a bit. I created three tiny ontologies.
One uses URNs. One uses example.org http URIs (thus non-
dereferencable) and one uses dereferencable URIs (though just Hash).
The test cases are located:
http://www.cs.man.ac.uk/~bparsia/2007/temp/hclsGenericToolsTests/
I started with the W3C RDF Validator <http://www.w3.org/RDF/Validator/>:
1. <http://www.w3.org/RDF/Validator/ARPServlet?URI=http%3A%2F%
2Fwww.cs.man.ac.uk%2F%7Ebparsia%2F2007%2Ftemp%2FhclsGenericToolsTests%
2Furn_testing.rdf&PARSE=Parse+URI%3A
+&TRIPLES_AND_GRAPH=PRINT_TRIPLES&FORMAT=PNG_EMBED>
2. <http://www.w3.org/RDF/Validator/ARPServlet?URI=http%3A%2F%
2Fwww.cs.man.ac.uk%2F%7Ebparsia%2F2007%2Ftemp%2FhclsGenericToolsTests%
2Fexample_org_test.rdf&PARSE=Parse+URI%3A
+&TRIPLES_AND_GRAPH=PRINT_TRIPLES&FORMAT=PNG_EMBED>
3. <http://www.w3.org/RDF/Validator/ARPServlet?URI=http%3A%2F
%2Fwww.cs.man.ac.uk%2F%7Ebparsia%2F2007%2Ftemp%2FhclsGenericToolsTests
%2Fexample_org_test.rdf&PARSE=Parse+URI%3A
+&TRIPLES_AND_GRAPH=PRINT_TRIPLES&FORMAT=PNG_EMBED>
In each case you get hyperlinks for the URIs, but you get a different
behavior if you try to click on them. In safari, the last example
downloads a file which really isn't what I want. I'll also note that
I think the built in vocabulary term links are broken too. I don't
like what's served at the other end of them. (Swoop in fact let you
override links to the rdf.rdf/owl.owl/etc. nonsense and instead let
you go to help pages about the term.)
I also used OWLSight:
http://pellet.owldl.com/owlsight/
(Warning: has crashing problems with Safari. It should warn you.)
Unfortunately, OWLSight doesn't give you URIs for loaded onts. But if
you load the URN ontology, OWLSight hyperlinks will navigate only in
the local context (i.e., inside the ontology). It has the same
behavior for all three ontologies (and is what I want for that
ontology). And would work for Hash vs. slash and for arbitrary
mixtures of URIs in the document.
Protege4alpha and Swoop hyperlinks behave the same way.
I installed Piggybank, but it doens't seem to work with the directory
or the raw RDF files, so I suppose I need to derive an html file or
something to see what happens (assistance here would be welcome).
I have no idea how to test Swoogle, or even what to test for.
Searching for a URNterm should be fine no matter how many documents
statements involving it are scattered across as long as it has those
documents in its index. It will do a bit better spidering if you use
resolvable http URIs for your classes, but you can get around that by
having seeAlso like links for various terms or term spaces. Indeed,
the problem of *finding* ontologies to collect seems rather different
from how you name the terms. I see *why* one might want to overload
these functions, but it doesn't seem like the workarounds are all
that painful. (For one, if you *import* the other ontology, it will
be spiderable.)
OWLSight also tests pellet, which seems to be indifferent to all
this, as I would expect. Importing these ontologies using the files'
uris will work normally and correctly in all tools. If you try to
import the urn, that won't work without a special resolver, but there
also is no URNed ontology to import.
Other tests or tools to test welcome.
Is this sort of investigation welcome? Useful? Welcome *and* useful? :)
Cheers,
Bijan.