[I have only read the past month or two of archives on the list; if I am re-treading a worn path, my apologies. But you did ask for comments by today …]
In a previous thread Elizabeth McKelvey correctly complains that there is a problem permitting both 6-digit extended years and dates without hyphens or times without colons:
EM> However, this option creates too much complexity. Systems will possibly have an extra format to parse. It also leads to possible data errors. How to remember it is okay to use 19990415 to represent year, month, and day, but not okay to use 199904 to represent a year and month?
Ray Denenberg responded that this is a tough nut to crack:
RD> I don't think there is any way around this problem though I would welcome further discussion.
RD> I do not think we could get agreement to profile out the separated forms (hyphens in dates and colons in times) because that's the form that W3CDTF uses and I don't think this spec is going to go anywhere if it does not accommodate W3CDTF.
And, without any doubt, the separated forms are easier for humans. It would be a really bad idea to profile them out.
RD> And on the other hand, part of the motivation for beginning the work on this spec was to add the non-separated forms.
A motivation I, for one, find very hard to understand. Many of us summarily stated in our various projects “always use the extended form” (which is what ISO 8601 calls the separated form) long before the W3C datatypes or profile came out, and for good reason. And unless I’m missing something, it’s pretty easy for legacy projects that wish to convert from unseparated to separated. (Just to prove to myself that it isn’t that tough, I wrote a toy converter that handles just the basic formats. You can find it at http://polo.services.brown.edu/staff/Syd_Bauman/temp/un2separatedTemporal.xslt.)
Permitting both separated and unseparated formats in a specification immediately doubles the work of anyone interested in implementing it, and in this case it would be for what, to me at least, is no gain and a little loss.
But at least as importantly, what is the motivation behind 6-digit years? I find it a bit hard to come up with a use case for normalized date representations of more than 4 digits at all, let alone for bibliographic metadata. Even if we were citing works written by the Legion of Superheroes in the 30th century, we’d only need 4 digits. And as for the past, writing wasn’t invented until the 4th millennium BCE. Even if you stretch it to include ideographic writing systems, you don’t need more than 4 digits.