On 1/10/2012 9:56 AM, John Gilmore wrote:
I too have some sympathy with Paul Gilmartin's objectives, but a
twos-complement representation of a hardware clock would be
problematic in many ways.
Dates alone are of course unproblematic. A signed fullword
Gregorian-Day value is usable to represent dates ±5.8 million years
about the Gregorian calendar's epoch origin, midnight, Sunday, 0000
December 31; a signed doubleword one would take us back to the Big
Bang.
Moreover, dates are enough for most events that occurred before AD
1900. Apart from those of certain astronomical events--Eclipses are
the obvious example--that can be calculated with great precision, the
time-of-day of these events is not usually known with any precision.
More on-topic is the fact that arithmetic with STCKE values is not
difficult. There are indeed contexts in which it is trivial.
Mathematicians who are interested in prime number sequences and
densities have had multiple-precision binary arithmetic packages
available to them since the days of the IBM 704.
Or again, because I write a lot of AMODE(64) code, I have written a
set of macros that do multiple-precision binary arithmetic using
signed-fullword arithmetic set-symbol elements; and there is reason to
believe that [too] many others have had to write such macros too.
I see "unproblematic" and "not difficult". I'm at a loss concerning
what you
do find "problematic". Perhaps the behavior of the clock comparator on
signed values?
Of course, I wrote that aware that you'd find any representation other
than integral day ordinal an inferior choice.
One comment on terminology is in order here too. In the light of the
substance of Swift's 'modest proposal' Paul Gilmartin's 'immodest
proposal' merits description as one.
If I'm indicted for the sort of naive etymological reversal that transforms
"university" to "multiversity", I must plead, "guilty".
-- gil
John Gilmore, Ashland, MA 01721 - USA
