Alexander Johannesen wrote:
>
> Seems some of my stuff on identity management has gone unanswered, so
> I've written a quick blog post to try to clarify why this is
> important, why this is how you should at least ponder identity in an
> RDF world, and on cataloging using nothing but URIs for both subject
> and resource identity ;
>
URIs are bad identifiers, for one reason: URIs always contain
verbal components, and words (names are words too) - sooner or later -
get changed. Also, the syntax of an URI might be valid for less than
eternity.
Least prone to changes are plain numbers.
Example:
Currently, one might be tempted to use WorldCat Identities for URIs.
Here is one:
http://orlabs.oclc.org/Identities/lccn-n79-3362
Just about anything in this, sooner or later, may have to be changed for
some compelling reason or other. The bare number: 79-3362, has the
best chance for longevity. If we store just the number, in an
appropriate field so software knows what it is, then software
can always supply the rest of the URI string following any current
fashion, and we don't have to change zillions of records when OCLC
decides to change their name or http is no longer the protocol to use.
The use of bare numbers has been the method in this country for a
long time. Everything else has changed, not the number. So, we had no
trouble with the records, we needed only change the software that makes
use of the numbers. For an equally long time, the MARC world has used
strings (names, phrases) as identifiers. As long as the strings don't
change, they are just as good identifiers as numbers. But some of them
got to be changed all the time, and that has meant big trouble for all
the databases that use them when they might have used the numbers
instead. But of course, numbers means one has to have access to the
authority data as well for indexing and/or access. That's why the MARC
world (or some of the vendors?) found it too difficult to use numbers.
And Alex: as long as we have to manage physical items in libraries,
and objects like documents, we need records to represent and describe
them. Or how else would you suggest to do that?
B.Eversberg
Received on Tue Oct 27 2009 - 05:59:31 EDT