On Mon, Jun 6, 2011 at 2:08 PM, Ray Denenberg
<[log in to unmask]> wrote:
> From: Bruce D'Arcus
>> What's the basis on which level two and three features are distributed?
>
> "are" distributed or "would be" distrubuted?  At this point the distribution is just a suggestion, and comments are solicited.

The latter. E.g. what's the basic on which you made the strawman distinctions?

> Level 0 is straightforward: anything reatained in the spec that conforms to ISO 8601. (And I propose to eliminate the distinction of whether or not it is part of W3CDTF.)  And so those features retained that are not part of ISO 8601 would be distributed over levels 2 and 3, and the question is on what basis.  I see three factors: 1. difficulty of implementation; 2. need; 3. separability.
>
> By "separability" I mean that I think we would want the BNF to be three (somewhat) distinct sections (perhaps all referencing a common part) corresponding to the three levels.  Placing a feature in level 2 vs. level 3 (or vice versa) might make it easier (or more difficult) to achieve the desired separability.
>
> I plan to begin looking at the BNF in terms of the above, but I would appreciate input from implementers and users about which features should be level 2 and which should be level 3.

For our usecase (bibliographic citation formatting, data exchange,
etc.; Zotero, CSL, Mendeley kinds of use cases), we would probably end
up cherry-picking from levels 2 and 3. We're unlikely to support
anything but very general uncertainty, for example, and not things
like masks.

One way to slice the levels is how different feature diverge from the
base. For example, a standard range of "2002-02/2002-03" is not
supported by Level 0, but it's components are.

OTOH, 2002-0u has no reasonable approximation I can see in ISO 8601.

Perhaps that suggests they should go on different levels?

I don't know; just a question.

Bruce