Some questions we're working on at the moment:

• It looks as though sources should have an "original date" field. However, it strikes me that the origDate may sometimes be as imprecise as the regular publication date, so perhaps we should add three new fields for the original publication date, giving us six date-related fields in the source:

  Date for the actual source:
      sc_NotBefore
      sc_NotAfter
      sc_DatePrecision

  Date for an original publication of the material:
      sc_origNotBefore
      sc_origNotAfter
      sc_origDatePrecision

  This is getting pretty complicated, though, and it might be overkill.
• It also strikes me that references come from sources, so the date fields in references may be superfluous; surely any reference inherits the date(s) of the source from which it comes? If so, we could eliminate them, which would balance the equation by simplifying the references table as we complicate the sources table.
• Finally, with regard to datePrecisions, I've been looking at the TEI XML specification, where there a mechanism for specifying precision, with the following values:

  "high" | "medium" | "low" | "unknown"

  This looks reasonable, except that we would also probably want to add "exact", or something similar, and make it the default value; so where a date is known to be correct, the default value would work. If you think about it, these values are not actually for precision; they're for certainty. notBefore and notAfter give us one type of precision (the range of possible dates), and the date format itself, in those two fields, gives us another -- you can do:

  1750
  1750-12
  1750-12-25

  i.e. year-level, month-level or day-level precision. So I wonder whether "datePrecision" ought really to be renamed "dateCertainty".