Bernhard Eversberg wrote:
<snip>A URI does not have to be a number--it is *any character string* that identifies a resource uniquely, and this includes textual strings as well. This coincides precisely with what our authority headings have been designed to do and I see no reason why we should not try to take advantage of this huge amount of work right now. So, for libraries that follow LCSH, the URI could just as easily be
URIs, just like textual strings, are subject to change although not meant to be. Bare IdNumbers are a little better (and much shorter). In most cases, URIs are all alike, and the only difference is an IdNumber contained in them.
So, why the trouble to store the entire URI with every record affected, when the number is all that is actually needed, and a changed URI most often differs not in the number but in some
other part. For example:
We might have
650 $u http://id.loc.gov/authorities/sh85090739
for the subject heading "Neo-Kantianism".
since sh85090739 is supposed to equal "Neo-Kantianism" (that is, if the catalogers have been doing their jobs competently) and consequently, there is no need for the nightmarish change of all our headings to numbers. This is how it works now in dbpedia with the URI using a textual string: http://dbpedia.org/resource/Neo-Kantianism. (Looks like the English abstract should be added to this record)
If this is the case, your suggestion for adding the "http://id.loc.gov/authorities/" as a separate function could work right now, today. The problem is that changes would have to be made at id.loc.gov to make it work as a real web service, so as to provide the XML that local catalogs could work with. As a simple illustration of how something similar works, see how I have implemented OCLC's web service see, e.g. http://www.galileo.aur.it/cgi-bin/koha/opac-detail.pl?bib=26352 and click on "Get a citation" in the right-hand column. There could be much better and "cooler" possibilities than this, however, for example, bringing in related information for the subjects.
What about those libraries that do not use LCSH but another controlled vocabulary? They could provide the same service for their headings and their catalogs, and then at the higher "LCSH-Other controlled vocabulary" level, related terms could be linked in some way similar to VIAF, or I am sure there are other ways as well. In dbpedia, you can see it in the Neo-Kantianism record (above) using the "owl:sameAs". This is how linked data can work, there could be owl:sameAs for all kinds of authority files, including dbpedia. Imagine the power of something like that.
Would this work 100%? Anything you decide to undertake will have problems, but it could provide at least 75-80% if not more, and could be done right now, with a minimum of cost and no disruption to cataloging productivity.