Re: Temporal database - no end date

>> Given there is physical limit to the smallest point of time we can record, constrained by planck length, I cannot see how the fact that time may be an infinitely divisible continuum has any relevancy to databases. <<

But is a continuum a good **model** for time? I am not going to have a database engine that can carry that many decimals of precision. I will be happy with the FIPS-127 Conformance test suite that wants to see at least 5 decimalls on the seconds. Now I can do math with my temporal model; rates like MPH or time series stats like MTF make sense and can be computed. If I have a (start, end) temporal pair, can I tell if a point in time falls inside or outside of it, even when that point is expressed to greater precision than I am using? Yep!

Is a quantum a good **model** for time? Well I cannot do rates and statistics with it. That is a killer issue right there.

Since durations do not exist in this model, I cannot really do much with data from a different source that are not on the same quantum model. Are you an SF fan? Ever read Philip Jose Farmer's Dayworld series? A future dystopia assigns the citizens a day of the week when they can live and freezes them the other six days. There are seven disjoint cultures that do not communicate, separated by a quantum with a granularity of seven days, but Tuesday does not exist for the Monday world. The plots revolve around a "day breaker" who lives multiple lives, blah, blah, blah.

I also don't think that progtammers who have not done scientifiic work realize how much a modern compiler does symbolic manipulations. In the old days when I was writing FORTRAN (not FORTRAN II, not FORTRAN IV, not Fortran 77 -- I am old!) the compilers would see SQRT(2.0) and spit out a floating point value immediately. Today, they hold the expression "SQRT(2.0)" and see if it gets used somewhere that allows a simplification, or if things can be re-arranged to reduce errors. For example, if you are adding up the numbers in an array, you do not have to add them in subscript order in spite of a loop that implies that.

With IEEE floats, we all kinds of symbols to play with -- +0 , -0, +8, -8, signaling NaNs and quiet NaNs! Received on Sat Jan 20 2007 - 21:58:13 CET