A comment on Jon Udell's Speaking and writing webscale identifiers.
Adding content hints to identifiers degrades into the case where you have several variations of the identifiers in the wild. For automation to help us it needs to know which identifiers are for the same thing. If all identifiers used the same syntax then it would be possible to automatically do this. Using the same syntax does not seem likely to happen given the number of global identification systems we already have, eg Handle (RFC 3650), DOI, URL, ISSN, ISBN, LOC, etc...
The core of the problem is not coming up with another identification system but coming up with a identification relationship system. This would address not only the same thing with multiple identifiers but also the relationships of parts to whole and variations, such as revisions or translations. Let's work on that for a while (independent of the larger semantic-web stuff).