Regarding "Don't quite see any benefit in having another set of
redirectable identifiers for the actual representations".
I have tried to explain this many times and I guess that I am just
not good at it. Let me try again, by asking you to review some
statements, and perhaps spend a few moments answering some questions
about them, and providing the reasoning behind your answers:
- http://purl.uniprot.org/uniprot/P12345 is out of date.
- http://purl.uniprot.org/uniprot/P12345 dc:creator "eric jain"
- What has changed about http://purl.uniprot.org/uniprot/P12345 in
the last year?
- Is http://purl.uniprot.org/uniprot/P12345 mentioned in the paper
identified by pubmed id 4030726?
- What sort of thing is http://purl.uniprot.org/uniprot/P12345?
- An organism of what species created http://purl.uniprot.org/uniprot/
P12345?
- An organism of what species created some instances of http://
purl.uniprot.org/uniprot/P12345?
- Can http://purl.uniprot.org/uniprot/P12345 be represented as a
sequence? A sequence of what?
Here are the questions that I would like if you could consider:
a) Does the statement make sense?
b) What does the statement mean?
c) By what algorithm should a (non human) semantic web agent
determine the answers to the above.
d) How would you represent the statement in RDF?
Now perhaps you don't want to make such statements, but I think it
would be presumptuous to think that others won't. As you work these,
perhaps you may notice that they would be much easier to answer had
they been phrased using "identifiers for the actual representations",
and if there was a way to know in which cases we were even talking
about a "representation" versus some stuff in the real world.
I suppose it might be possible to represent which header should be
used in the content negotiation as part of the RDF, but a) It's got
to be easier to just put that information in the name and b) In the
case that you want to, e.g. mirror some contents of Uniprot on a file
system, you will have to make up distinct names anyways? Maybe I'm
dense, but I fail to see how content negotiation is of any use on the
semantic web.
The cost to a provider, particularly one as important as Uniprot, to
make it possible for others to unambiguously make statements, is
extremely small compared to the cost that is and will be devoted to
figuring out how to navigate the semantic web if it doesn't. It is
cheap to provide distinct identifiers for the various things are
referred to, and (in many cases) only slightly more expensive to
provide some clarity about what these identifiers denote.
-Alan
ps. LSIDs vs URI's aside, it's admirable (and greatly appreciated)
that Uniprot is on the way to providing stable identifiers. I do hope
you will be continuing in this effort to the bitter end:) For
instance on the html page, we have URLs such as http://srs.ebi.ac.uk/
srsbin/cgi-bin/wgetz?-e+[HSSP-ID:'7aat'] which could be profitably
hidden behind a redirect using the same mechanism that you are
already using.
On Jul 12, 2007, at 11:56 AM, Eric Jain wrote:
Xiaoshu Wang wrote:
IMHO, I think it would be nicer and less confusing if you make
"http://purl.uniprot.org/uniprot/P12345"; a skeleton and 303
redirect to either "http://purl.uniprot.org/uniprot/P12345.html";
or "http://purl.uniprot.org/uniprot/P12345.rdf"; depends on the
value of Accept header.
That's what I am doing?
e.g. try
`curl -L -H "Accept: text/html" -v http://purl.uniprot.org/uniprot/
P12345`
`curl -L -H "Accept: application/rdf+xml" -v http://
purl.uniprot.org/uniprot/P12345`
...except that I redirect to:
http://beta.uniprot.org/uniprot/P12345 or
http://beta.uniprot.org/uniprot/P12345.rdf
(Don't quite see any benefit in having another set of redirectable
identifiers for the actual representations.)
