Bruce (and everyone).  Here are some thoughts I have on all of this.

First of all I think it may have been a mistake to write the BNF in a manner such that  "imprecision"  is  pervasive rather than isolated.  (And I'm using the term "imprecision" as a generalization of "questionable" and "approximate". It might not be the best term and suggested alternatives are welcome.)

In other words, we have:

baseYear = digit digit digit digit
year = baseYear | "("baseYear")" ("?" | "~")

And so everywhere "year" is referenced it picks up all the imprecision baggage. (And similarly for day, month, etc.)

Instead, an approach along the lines of:

Year = digit digit digit digit
impreciseYear = Year | "("Year")" ("?" | "~")

And isolate everything that has to do with "imprecision" in one section.

And further, I think it is reasonable to consider having two (or more) conformance levels, where at the base level, imprecision need not be supported.

I also think that instead of

("?" | "~")

it could be

("?" | "~" | "~?")

That is, you can represent a date as "questionable", "approximate", or "approximate and questionable"

Where "approximate and questionable" is self-explanatory and needs no further definion.

Complex constructs like

((2001)?-01)~

Simply would not be respresented.

I would like feedback on these suggestions, before proceeding.

--Ray

> -----Original Message-----
> From: Discussion of the Developing Date/Time Standards
> Sent: Sunday, May 22, 2011 4:26 PM
> Subject: Re: [DATETIME] New approach about question marks - a
> suggestion
>
> On Sun, 22 May 2011 16:05:01 -0400
> >On Sun, May 22, 2011 at 3:43 PM, Markus Flatscher
> [...]
> >> Not to stretch things---just to make sure we're on the same page:
> you still intend to distinguish precision ("approximate") and certainty
> ("questionable"), right?
> >
> >I believe that's the intention. But I have to say, the distinction
> >remains entirely opaque to me. What's the real difference between
> >"2000-01-01?" and "2000-01-01~"? To me they both seem to say "a day
> >somewhere around January 1, 2000; not exactly sure though."
> >
> >I would hope the spec can clear this up.
>
> This is how I see the distinction, maybe this helps:
>
> (1) The natural-language surface phenomenon that reflects this
> distinction is "ca." (precision/"approximate") vs. "?"
> (certainty/"questionable").
> (2) I can be _certain_ that a date is imprecise (approximate: "ca. 1
> March 1999"), or I can be uncertain about whether the date is correct
> ("1 March 1999?").
>
> Sidenote: It's a distinction that feels natural to people in text
> encoding, since it applies generally, not just to dates. Cf. TEI P5
> Guidelines, section 21.2 "Indications of Precision" (http://www.tei-
> c.org/release/doc/tei-p5-doc/en/html/CE.html#CEPREC)
>
> Markus
>
> --
> Markus Flatscher, Editorial and Technical Specialist ROTUNDA, The
> University of Virginia Press PO Box 801079, Charlottesville, VA 22904-
> 4318 USA
> Courier: 310 Old Ivy Way, Suite 302, Charlottesville VA 22903