Having read the discussion about whether we are developing a format for "user input" I wish to weigh in.
However first let me briefly address the issue of "dual representation" which is where this started, and which I think needs to be treated separately.
'Most obviously, in most data interchange applications, it's usually regarded as preferable to eliminate as much variation as possible in formats: if there is no difference in meaning between "2011-01-21" and "20110121", people interested solely in the exchange of data among databases will (again, in my experience) unanimously prefer to choose one of these formats and require it, rather than allowing either. That seems good practice to me.'
Indeed it is in general a good practice. The problem comes when you have two different constituencies that make compelling cases for the two different forms. But let's leave this for a different thread.
Back to "user input". I wish to highlight the distinction between "format for user input" and "human readable format". I believe we are NOT developing a format for user input, and that we ARE developing a human readable format - with the following caveat: "'human readable' is in the eyes of the human." And along with "human readable" goes "human typeable".
Bear with me, or skip the following two paragraphs if you like.
I have spent a number of years working on the CQL query language. http://www.loc.gov/standards/sru/specs/cql.html. Debate has raged among CQL developers and implementers whether it is intended as a human readable language. And the answer (my answer, anyway) is that it is intended to support very simple to very complex queries, and the level of complexity of a given query string is comensurate with the complexity of the query. So simple queries, like "cat and dog" are human readable/typeable. But there are complex queries supported that someone well versed in CQL might be able to read but most users would not, and nobody would want to type them.
A number of years ago, an information retrieval scientist suggested that there be a committee to develop a "user input" standard for CQL. It would be an adjunct to the CQL standard, not part of the standard itself: a format for user input that would make it easy for a user to create complex queries without knowing the actual CQL syntax. This "user input" format would not be an interchange format for tranfer across systems, it would be converted at the client system to CQL before transfer. Its purpose would be to give a client developer a standard syntax on which to build a user interface. Long story short, this effort never got off the ground. My point is that this is what I think of as a "standard for user input".
Bottom line is, we are not developing a format for user input. Whether we are developing a human readable format is a different question and I think the answer is that it is human readable to the extent that ISO 8601 is human readable, which (as I said above) is in the eyes of the human.