He reminds people of the three layers in the classical library metadata stack: encoding (ISO 2709 or Z39.2), content designation (as expressed in the various MARC formats), and content values (which is the focus of cataloging rules and controlled terminologies). I was interested to see this, as I have been emphasising the importance of these distinctions in recent presentations (see, for example, [ppt] and [ppt] for two related presentations).
Dublin Core took a different approach. It initially focused on the 'designation' part of the stack (what are the elements of importance), and latterly produced several encoding recommendations. However, there are no generally deployed content standards associated with the Dublin Core: it did not set out to develop any, leaving this to community agreement. And, indeed, several such agreements have arisen. Similarly, other metadata initiatives - IEEE LOM, for example - have not focused to the same extent on content standards.
This then plays into one of the major experiences of the initial harvesting projects: they soon discovered that the absence of content standards creates interoperability issues.
I mention content standards, because I was surprised that consideration of AACR3 was not brought more into the xml4lib discussion. Based on recent conversations with colleagues within and without OCLC, I think that two issues are vital as AACR3 is worked through:
- The majority of data is going to be consumed by programs: it needs to be designed in such a way that it helps rather than hinders this process.
- The value of existing approaches is being questioned, especially in the full glare of the Amazoogle world. This suggests that a clear focus on cost and benefit is important in revising the code: valuable cataloger time should be spent where it creates most value.