 |
 |
Hi Karen: On Wed, Jan 22, 2014 at 3:16 PM, Karen Coyle <[log in to unmask]> wrote: > I can't address the first points, but I can speak a bit to the question of > meaningful URIs. In the original creation of the RDA elements, "meaningful" > URIs were used based on the actual RDA terminology. This resulted in URIs > like: > > http://rdvocab.info/Elements/alternativeChronologicalDesignationOfLastIssueOrPartOfSequence > > and... > > http://rdvocab.info/Elements/alternativeChronologicalDesignationOfLastIssueOrPartOfSequenceManifestation > > Not only that, the terminology for some elements changed over time, which in > some cases meant deprecating a property that was then overly confusing based > on its name. > > Now, I agree that one possibility would have been for the JSC to develop > meaningful but reasonably short property names. Another possibility is that > we cease looking at URIs and begin to work with labels, since URIs are for > machines and labels are for humans. Unfortunately, much RDF software still > expects you to work with the underlying URI rather than the human-facing > label. We need to get through that stage as quickly as possible, because > it's causing us to put effort into URI "naming" that would be best used for > other analysis activities. Thanks for responding on this front. I understand that, while the vocabulary was in heavy active development it might have been painful to adjust as elements changed, but given that this marks the actual publication of the vocabulary, that churn should have settled down, and then this part of the JSC's contribution to semantic web could have semantics applied at both the micro and macro level. I guess I see URIs as roughly parallel to API names; as long as humans are assembling programs, we're likely to benefit from having meaningful (no air quotes required) names... even if sometimes the meaning drifts over time and the code & APIs need to be refactored. Dealing with sequentially numbered alphanumeric identifiers reminds me rather painfully of MARC. For what it's worth (and it might not be worth much) "curl http://rdaregistry.info/Elements/a/P50101 | grep "reg:name" | sort | uniq -c" shows that the reg:name property is unique across all of the agent properties, at least. Remnants of the earlier naming effort? If that pattern holds, those could have been simply used for the identifiers in place of "P#####". The most unwieldy of those appears to be "otherDesignationAssociatedWithTheCorporateBody" (which _is_ unwieldy, certainly, but still more meaningful than http://rdaregistry.info/Elements/a/P50033). Perhaps it's not too late?