On Wed, 13 Jul 2011 17:59:58 +0200, [UTF-8?]SaaÅ¡ha Metsärantala wrote
> > Again.. we are back to p-r-e-c-i-s-i-o-n... ! :)
> Precision and reliability seem to have been carefully disregarded by (most
> of) earlier specifications. An assumption of total precision makes things
Precision (readability and repeatability) is something quite different from
reliability (certainty, belief, trust). Its yard-stricks versus data
quality. Measures of reliabiliy are, of course, built upon models of
precision but are also fundamentaly qualatative rather than quantative.
> easier. But it is probably just there, work need to be done.
> Our discussions may seem lengthy, but this comes as no surprise. It is a
> difficult issue. Let's work a little bit more with that: It seems that we
> will soon reach some good way to formulate a really useful EDTF
> > the data elements definition would dictate which interpretation applies.
> For interoperability reasons, I would really like to limit the number of
> interpretations for each syntactic construct. I do NOT mean that we should
> not be aware of the different possible interpretations! Let's instead
> augment the number of syntactic constructs whenever there is a need for
> that and document their semantics in great detail.
> If there are three ways to interpret an interval, then let's describe
> three different syntactic constructs and their semantics.
> Over-simplification seldom solves complexity problems.
No. There are not multiple ways to interpret an interval. We are, I think,
all quite clear what closed intervals are: a continuum of points between two
discrete (end)points. Where these have room is in their application. This,
however, is the same for dates (irrespective of precision). This is beyond
the scope of our work but something for those developing the dataset
profiles using our work.
Edward C. Zimmermann, NONMONOTONIC LAB