Note that here I gave the same GUID to all three representations of the single publication. It implies that the GUID is ID for the literature object, not for (meta) data objects. If we need GUID for data objects also, we would need two types ofGUID.
I would encourage everyone to have a look at the report from the recent TDWG/GBIF Workshop on GUIDs:
http://wiki.gbif.org/guidwiki/wikka.php?wakka=GUID1Report
One of the outcomes of this workshop was the need to clarify the distinction between GUIDs assigned to data objecs, vs. GUIDs assigned to conceptual entities. I was a member of a breakout group that discussed this distinction, and we were tasked with creating a page or set of pages on the Wiki to clarify this distinction -- particularly in the context of LSIDs (the GUID technology which seemed to be strongly preferred by the workshop participants). It's a subtle distinction with very significant implications (it's also a topic muddled with unfortunate semantics). Our breakout group concluded that GUIDs for conceptual entities serve an important role in biodiversity informatics, if for nothing else than to serve as a "hub" around which multiple data-object GUIDs may cross-reference each other.
So...James...in your example, there would probably be multiple data-object GUIDs for the different digital "renderings"; and perhaps a separate data-less GUID assigned to the the "concept" of the specific publication instance, to which all of the alternative data-GUIDs may be "hubbed".
This group should also take note of point 5 under the "Work plan" section of the Workshop Report, regarding the need for a sort of "Publication Bank".
I meant ISO dates only for "date published (as corrected)". I agree with you date issue; in Level 1, date published as cited should be as is, shouldn't be interpreted.
But wouldn't we also want the "corrected" (="interpreted") date (or date range) to be among the attributes -- even at level 1? If for no other reason than estimating chronology?
Aloha, Rich