Print

Print


Ok, I think I have a solution to all of this.

'u' clearly has a much different usage when used internally than when it
replaces one or more consecutive rightmost digits. I don't think we have yet
confronted this.

Thus 196u means:
    "a date in the 1960s"  

and expresses imprecision, where 1u66 might be said to express uncertainty.

In other words I suspect the solution to this problem is not to refer to
"rightmost" 'u' as "uncertainly" and instead refer to it as "imprecision". 

And since 'u' means different things when used internally vs. rightmost, and
perhaps 'u' shouldn't be used for both. 

I therefore propose that 'x' be used instead of 'u' for level 1 uncertainty,
and that we rename "uncertainly" as "imprecision" for level 1 (and continue
to call it 'uncertainty' for level 2 - internal).

--Ray


> -----Original Message-----
> From: Discussion of the Developing Date/Time Standards
> [mailto:[log in to unmask]] On Behalf Of Edward C. Zimmermann
> Sent: Tuesday, August 02, 2011 2:26 PM
> To: [log in to unmask]
> Subject: Re: [DATETIME] Rethinking Precision and "x"
> 
> On Tue, 2 Aug 2011 10:08:40 -0400, Ray Denenberg wrote
> > From: Edward C. Zimmermann
> > > The same "problem" exists with date expressions. We have well
> > > defined expressions of year, month, day, hour, minute, second,..
> > > perhaps even century precision.. but not decade, whence the
> generalized "x" syntax.
> >
> > Ed - I appreciate your patient efforts to enlighten us on precision.
> I
> > for one have been a bit slow to grasp your philosophy, particularly
> > the role of the "x" syntax, but the above passage has provided some
> > clarity and has motivated me to think this through more clearly (I
> hope).
> >
> > So, If '1965' means "some date  - or point in time - during the year
> > 1965",
> 
> Its a bit more philosophical.. er... ambiguous. A point can also be an
> interval.. and an interval at a coarser precision can be a point...
> When, for example, was the WTC Tower 1 (aka North Tower) destroyed?
> When did WW-II start? When did it end? When did the "Berlin Wall" fall?
> When did the "Cold War" start? Or.. how about the time of an explosion
> of a stick of TNT--- think about the velocity and wave?
> All events have a start and an end.. BUT.. OK.. that is getting a bit
> too ...
> 
> 
> > then '196x' means "some date -  or point in time - during the 1960s".
> >
> > Right?
> 
> Its means a date (leave off "some" since its too loaded) in the 1960s.
> From the perspective of decade measurement precision its a point in
> time..
> 
> >
> > That's really not the way it is represented in the current draft.
> > Rather,
> it
> > is included within multiple dates. Based on my new (and hopefully
> > correct) understanding I don't think it belongs either in multiple
> > dates or one of a set.
> >
> > But let's digress momentarily and revisit the 'u' syntax.  The draft
> spec
> > currently says '199u' means "some unspecified year in the 1990s".
> But do
> 
> While u can have a measurement semantics.. its more .. its both an
> explicit declartion of "unknown" or "uncertainty" AND a placeholder to
> one day get filled with the a known digit.
> 
> Example:  19u4-12-12
> 
> This is saying that an event (measured in day precision) occured on the
> 12th of December in the 4th year of some, as yet unknown/unconfirmed,
> decade in the 20th century.
> 
> 19uu-12-25
> An event took place on one, as yet unknown, Christmas day in the 100
> year range of 1900 to 1999.
> 
> When the "missing digit(s)" is one day disclosed there is little to
> indicate that the rest will change. We are in fact saying that it
> won't--- at least not from the perspective of the state of knowledge
> one had when one expressed the date.
> With 'x' or the other precision expressions we are saying.. Yes.. it
> can change..  I wrote about knowledge and measurement a short while
> back...
> 
> A date expression 195x might become with a higher level of measurement
> precision 1960 or 1949.. 195u should not.. again.. think about weights..
> a gramm balance that measures to 1 gram precision.. A sample can have 1
> gram weight and yet with a lab balance that can measure to fractional
> mg it might have only 0.99001 grams.. It other words.. less than 1
> gram.. Imagine timing events with a Sun dial, sand hour, cheap LCD
> watch, an atomic clock...
> 
> 
> > we really want it to mean that or do we want it to mean: "some
> > unspecified date - or point in time - during the 1990s"?
> >
> > And if so, then wouldn't 199x mean the same thing as 199u?
> 
> It can.. But what would, by contrast, 19xx-12-25 mean?
> u is a placeholder.. x is not..
> 
> We can have 1u2u or even uu99 but not 1x2x or xx99. These, like, 19xx-
> 12-25 make no sense.  19xx is "100 year" precision.. 12-25 is day
> precision.. Its like saying.. I know how many rest micrograms a sample
> has but can only roughly estimate how many kilograms it has...
> 
> 
> >From a programmer's prespective.. matching dates with u is simple
> globing..
> like '?'. With x its not glob but comparing to least precision...
> 
> >
> > --Ray
> 
> 
> --
> 
> Edward C. Zimmermann, NONMONOTONIC LAB
> http://www.nonmonotonic.net
> Umsatz-St-ID: DE130492967