Re: Intervals (Was: NIAM (was: Just one more anecdote))
Date: 12 Aug 2005 05:47:28 -0700
Message-ID: <1123850848.774470.233330_at_g43g2000cwa.googlegroups.com>
>> Declaring column "MYDATE rdate" actually generated two columns, MYDATE_BEG and MYDATE_END. <<
I disagree. The nature of time is a continuum -- see Zeno and Einstein for details. Therefore a single point or a set of distrinct points (Chronons) just does not work. The ISO model of time and just about all the temporal database research uses half-open intervals, not Chronons. Date's book failed to get enough sales inspite of his name and it generated some "bad feelings" in the academic community when he attacked Rick Snodgrass who has done temporal RDBMS for 20+ years.
To be atomic, a data element has to have meaning only when it is whole. Extract a part and the meaning is lost. Example, the month taken out of a date is an infinite set of disjoint durations while the date is a single duration. Example: (longitude, latitude) marks a single (scalar) point on the globe, while either one is a line with an infinite number of points. The fact that SQL make me use one column, two columns or whatever is a syntax issue, not a modeling issue.
I have an article on this at DBAZine.com if you want to browse it. Received on Fri Aug 12 2005 - 14:47:28 CEST
