Charles,

        Very nicely put.  You re-surface a case that has shown up before but is worth reminding ourselves of -- the need to exress a range where one or both start and end termini can themselves only be recorded with no better precision than a date range.  This is a relatively common occurrence with cultural artifacts. ISO 8601 fails miserably on such a use case.

        The new variation you introduce is associating one or more termini with a named period which will have a range, but the boundaries of the range is subject to change as scholars refine their view.  Hmmm...sounds like RDF solution might be needed?

Ta,

Eric



-----Original Message-----
From: Charles McCathieNevile [mailto:[log in to unmask]]
Sent: Tuesday, January 11, 2005 5:45 AM
To: Childress,Eric
Cc: [log in to unmask]
Subject: Ranges as a terminus... Re: DCMI Date WG workplan 2004/2005 --
proposed


On Mon, 10 Jan 2005, Childress,Eric wrote:

>g. Soft termini (i.e. the outer bounds for one or more termini is known or
>can be associated with a known period, but one or both of the exact
>boundaries of the event referenced are not known)

>-- I suggest this approach because this issue is concievably a sub-category
>of most of the other categories (or perhaps arguably an inherent feature of
>most large dates) ; rather than add "and soft termini" or something similar
>to the various date categories, maybe it's simply easier to recognize the
>issue as a category in its own right?)
...
>It strikes me that we may need to begin compiling use cases for each of the
>categories?  If the WG agrees, I'll likely return to the WG to solicit a
>volunteer to act as use case master.

I think it is worth recognising this as an issue. And I think use cases are
very important.

One is in historical work, where over time one refines the information about
a particular object or document. For example, on their discovery one might
have dated the Dead Sea Scrolls as being compiled between some absolute date,
and some event which was in fact itself only roughly dated.

We *could* do this by modelling times as ranges automatically, where a
specified single time (the Hejira year 1473, for example) actually means the
largest valid range covered by that time - in other words, from the beginning
of 1 Awal Muharam wherever it occurs first, until the last day of
Dhou-al-hijja wherever it occurs last, and the gregorian date-time
2005-01-11T11:38+0100 as the range starting at the beginning of the first
second of that minute, and ending with the last second of it. (This already
includes time-zone information, which has a relatively simple model
specified with weel-understood encodings, something that might not be the
case for Hejira calendars which in any case are more complex).

For most purposes, I think being given a single time doesn't need any more
processing, since the fact that it is a range isn't important to the use
case. But it would be worth specifying (as XML schema does) that we are
already talking about something with a start and end.

Thus we could model all "dates" (including times, for people who want them)
as a thing with a start and end, specifiable themselves as a time (and thus a
range). In the simple case there is something like a gregorian date whose
termini are understood from the encoding, in the medium case there are two
such dates, and in the complex case one or other terminus can itself be
specified as a time range.

(This will make it easier to say that something falls between the Tang
dynasty and the wandering in the desert of the Israelites, as soon as we have
a reasonable encoding for those...)

cheers

Chaals