I would like to try to sum up what I've seen to date, and derived from the various discussion. Apologies for the long post. (Perhaps the CF leaders want to weigh in on whether they wish this conversation to continue on this list, or would like to promote it to another.)
I have begun collating material that is relevant at the following URL:
http://marinemetadata.org/examples/mmihostedwork/ontologieswork/guidesOntologyProvider/urirefs
(short link, to be created soon (Luis help?) is
http://marinemetadata.org/urirefs).
I would be happy to have you contribute material to this -- contact Luis for permissions to add such information -- and we will try to keep up with the multiple mail list references as well.
I support most of Luis' point re URLs. If we're having this discussion here, I urge we remove the '#' mechanism, which is broken from a service provider standpoint. (Because the server never sees the '#air_density' part of the URL, so many potential features are awkward or no longer possible.) Replacing the # with / is one alternative that seems functional.
After discussion with Roy, I concur that the non-semantic URIs are very important, but I think are not required in all cases. The design criteria is whether the semantic concept is permanently associated with that precise term. I see two situations when the association between concept and semantic term (URI) is permanent:
1) The concept is in fact referenced by the term -- consider a common word in a language, which at any given time has one (or many) concepts associated with it. While those concepts may change, the word itself is the lookup, and will always be the lookup. A reference dictionary for a language has this characteristic, even if differences over time must be reflected, the reference remains permanent.
2) The concept never changes, as in the example of Jonathon Gregory (of a static data set). As with taxonomies, a name may only be persistent within a given context, e.g., one data set or a particular time period. For dynamic contexts, an old name (no longer used) might overlap with a new name, matching a different concept. Therefore additional versioning information (a time period, or a vocabulary version) must be added to fully identify which term/concept is being referenced.
So a permanent URI may have two flavors:
A) A 'permanent reference' URI, with underlying definitions that may change over time, but 'air_density' will always point to one place, even as its content (definition) may change.
B) A 'permanent concept' URI, which will never change in key or in referenced concept. This URL can be made permanently unique through the (typically semantic) measures like one of
http://www.cfconventions.org/standard_name/20070318/air_density
http://wwwcfconventions.org/standard_name/V3Rev14/air_density
http://www.cfconventions.org/standard_name/WOCE2001/air_density
http://www.cfconventions.org/standard_name/15fc0c/air_density
depending on the versioning mode the provider wants to use. This is what Jonathon references, I believe.
As Roy notes, the problem with these is that someday/somewhere, 'air_density' (or 'WOCE2001', for that matter) may imply something different to the viewer, and so this URL may become misleading, even though it may be versioned and still literally accurate. In that case a less semantic URL like
http://www.cfconventions.org/standard_name/a345c26
or even less semantically, using a URN:
urn:org:urnregistry:a345c2615fc0c71c0c
will be better from a semantically neutral standpoint.
Of course, whether or not the served content or presentation behind any such permanent URI in fact changes will be a function of the policy of the owner of the services, if any, that are found via that URI.
John
>At 4:09 PM -0700 3/21/07, Luis Bermudez wrote:
>So maybe this is better:
>
>http://www.cfconventions.org/standard_name#air_density
>http://www.cfconventions.org/canonical_units#kg.m-3
At 7:59 AM +0000 3/22/07, Jonathan Gregory wrote:
> Because they exist in data which has been archived, we cannot abolish them. When we have revised our naming scheme, we create aliases for old names to match them to new names, so the old ones remain valid.
At 7:47 AM +0000 3/22/07, Roy Lowry wrote:
>I would much prefer semantically neutral keys to be assigned rather than using the text of the standard name in the URI. Standard Name texts have been revised in the past and there will probably be future changes, which destroys the permanence of the URI at a stroke.
--
----------
John Graybeal <mailto:graybeal at mbari.org> -- 831-775-1956
Monterey Bay Aquarium Research Institute
Marine Metadata Initiative: http://marinemetadata.org || Shore Side Data System: http://www.mbari.org/ssds
Received on Thu Mar 22 2007 - 11:04:03 GMT
