Re: Software Engineering News Brief

[eggert@twinsun.com (Paul Eggert)]

dewar@merv.cs.nyu.edu (Robert Dewar) writes:

> But to suggest that the standard built-in package have this kind of shaky
> semantics, i.e. semantics that so obviously conflicts with one reasonable
> expectation

If the Ada standard couldn't specify semantics that obviously conflict
with ``one reasonable expectation,'' then there wouldn't be a standard.
The whole point of a standard is to specify one among several reasonable
solutions to a problem.  All I'm saying is that for calendars, the most
reasonable solution is the Gregorian calendar.

I agree that dates are not as simple as you think.  But I'm afraid that
a good deal of the calendrical misinformation in this thread has been posted
by you: ``all countries agree for the Ada range of dates'' (they don't),
``the system of leap years has changed since [Pope Gregory's scheme]''
(it hasn't).

Perhaps the Ada standardization committee chose the years 1901-2099 based
on similar misunderstandings.  That wouldn't be too surprising,
as few computer language experts are also experts about timekeeping.
It reminds me of the C standard's bug in leap second handling:
the C standard allows for two leap seconds in the same minute
(an impossibility), because the C committee misunderstood leap seconds.

In any case, I urge the Ada standard committee to contact a genuine calendar
implementation expert before making any future decisions in this area.
A good choice would be Professor Ed Reingold of the University of Illinois,
coauthor of ``Calendrical Calculations'', Software--Practice & Experience
20, 9 (1990), 899-928.