Hello.
My name is Jon Holmen. I have just become a member of this list and
this is a response to macroy
about the fuzzy date format proposal. I think this is an interesting
work, but I also believe that before one can agree on a standard for
representing fuzzy dates, one has to clarify the underlying model.
This is an attempt to clarify at least some important aspects about
dating based on CIDOC-CRM ( ISO 217727:2006, standard for
documentation of museum material). I have contributed to a paper about
these questions presented at CAA2009, Williamsburg. The paper is
available here:
www.edd.uio.no/artiklar/arkeologi/holmen_ore_caa2009.pdf
Hera are some main aspects about dating seen from my point of view:
1.When you assign dates to something, you assign dates to one or a
collection of events or activities. F.i. if you assign dates an
object, you really assign dates to when it has been produced, used,
found or other; that is when, on the time line, one (or more) of these
events took place.
2.Any physical or mental event on this time line has duration. In
other words; our dates is intervals, not points. This is true even if
we often gives just one number for the representation. When resolution
is low enough this number covers both start and end of the given
interval. F.i. When you say «the treaty was signed at 08.05.1945», you
really say «it was signed at sometimes in between 00:00:00 08.05.1945
and 23:59:59 08.05.1945 (closed interval) and it took some time less
than a day» ( if you have chosen secs as your time resolution and we
do not consider time zones).
In CIDOC-CRM these dates are called timespans that is connected to
their respective events.
3.The fuzziness of the dating is reflected in the determination of the
start- and endpoints of these timespans. If we go back to the signing
of the treaty in pt.2, there are two questions.
The first is of course to define what we mean by signing the treaty:
Is it the whole ceremony, only the putting of signatures on the
document, or perhaps should some of the preparations also be included.
This is a question of what you are dating and should not be mixed up
with the actual model for the timespan, that has a clear cut
startpoint and a clear cut endpoint.
The second question is how to represent fuzziness in the measurement
of this (clear cut )start- and endpoint of the timespan.
Let us just say that in our example we are dating the putting of
signatures on the document.
Fortunately we have a photo (with a timestamp that we trust ) of the
signing, taken 12:15, 08.05.1945. And we have additional information
about the ceremony stating that it started 11:00 and ended 13:00.
In this case we would say that the signing took place at sometimes in
between 11:00:00 and 13:00:00, 08.05.1945. In CIDOC-CRM the relation
between the interval formed by these to dates and the timespan is
called «occurred within».
We also know that around 12:15 they were writing their names on the
document, so the start of writing their names had to be before, and
the end had to be after picture was taken. But since the resolution of
this dating is minutes, we can only determine the start to be before
12:15:59 and the end to be after 12:15:00. This gives us two
intervals, one for the occurrence of the start of the event (between
11:00:00 and 12:15:59) , and one for the occurrence of the end of the
event (between 12:15:00 and 13:00:00). As you can see, the dating of
the event with fuzziness needs 4 numbers (+ probably resolution,time
zone a.s.o.):
Earliest possible start, latest possible start, earliest possible end
and latest possible end.
In this example (so far) latest possible start is after earliest
possible end, leaving no sure point where the event actually took
place.
Luckily for us another photo shows up, still showing signing of the
same documents. This photo has timestamp 12:20. Adding this
information moves the earliest possible end to 12:20:00. So we have a
positive time gap between latest possible start (12:15:59) and
earliest possible end (12:20:00), and therefor an interval where the
signing was surely ongoing.
In CIDOC-CRM the relation between the interval formed by these two
dates and the timespan is called «ongoing throughout».
4.By using this conceptual «format» for dating, all kind of date
arithmetic in terms of relations between timespans, like occurred
before, overlaps a.s.o. (the so called Allen operators) is easily
extended from intervals with exact start and end to intervals with
fuzzy start and end. (See the paper).
I hope that this will clarify some of the aspects with fuzzy dates and
perhaps give more input to possible implementations of formats and
algorithms to work with these questions.
Best regards, Jon