Part of the difficulty in maintaining parallel registrations is that the
registration process is different and thus the registrations will get
out of sync.
If you're ever going to do print-by-reference through a URL, you'll need
to know the MIME types of that the printer accepts (and not its enums),
in order to properly set the "accept:" headers on the request for the
referenced data.
If you're ever going to allow a user to examine a job they've submitted
in the past, you'll need to know the MIME type of the job data, so that
you can properly set the Content-Type.
-- http://www.parc.xerox.com/masinter