## DATETIME@C4VLPLISTSERV01.LOC.GOV

#### View:

 Message: [ First | Previous | Next | Last ] By Topic: [ First | Previous | Next | Last ] By Author: [ First | Previous | Next | Last ] Font: Proportional Font

Subject:

Re: interval precision

From:

Date:

Wed, 20 Jul 2011 11:35:57 +0200

Content-Type:

text/plain

Parts/Attachments:

 text/plain (62 lines)
 ```On Tue, 19 Jul 2011 14:31:26 -0400, Ray Denenberg, Library of Congress wrote > We should avoid over-engineering. There is no need for a syntactic construct to indicate precision. > > The precision of an interval is really not a concept that needs articulation. If you have the interval: > > 1960/9164-04 > > Then the start endpoint and end endpoint have their own precisions - year and month precision respectively. These need not be the same, and the precision of the interval is simply undefined. > > --Ray An astute observation. It can't be defined. [sidetrack: The interesting problem is that an abstract time interval is not really a set of discrete points of undividable precision but a fuzzy continuum with measurements always at some precision. If we judge time considering precision we have a clear well defined element count for any interval and have a well defined set of points. The highest currently agreed upon ("interoperable") resolution of time is 1/9192631770 of a second (from the state transition of Caesium-133). If one second can have at most 9,192,631,770 sub-points then 1 day can have at most 24*60*60*9,192,631,770 or 794243384928000 elements-- by current computational standards a relatively small number (fits in 64-bit ints).] Does it matter if its defined? For, at least my, typical questions I don't think so. For any time point (expressed in any precision) I can answer the question: "Is it within the given interval?" Does "1960/1964-04" contain - "1963"? Yes - "1964"? Yes (since the precision of comparison can't ever be finer than the the precision of the least precise measure). - "1964-04-12"? Yes - "1964-05"? No Other typical questions such as: - Do the intervals overlap? - Is one interval included in the other? - Are the two intervals exclusive to one another (and, if so, their order)? similarly follow.... With a higher level of (measurement) precision, of course, it might turn out that the date expressed as "1964" in the comparison was really "1964-05- 12T12:30" and this NOT in the interval but is it really the same comparison? The state of knowledge, afteral, is different. The same observation, of course, holds for intervals with the same precision for start and end when we increase precision. Increasing the precision--- just as increasing knowledge--- can sometimes lead to new and different conclusions.... -- Edward C. Zimmermann, NONMONOTONIC LAB http://www.nonmonotonic.net Umsatz-St-ID: DE130492967 ```