Re: Software Engineering News Brief

[dewar@merv.cs.nyu.edu (Robert Dewar)]

Paul Eggert says

"> To accept dates before the 18th century without comment or error,
> and then misinterpret them seems highly error prone.

On the contrary, I've helped write some applications where quietly
accepting pre-18th-century dates was precisely the right thing to do.
A calendar package should not presume to tell applications what the
proper range is -- it should support the calendar,
and let applications worry about data validity checking."



An exactly wrong argument, at least for the Ada design philosophy. You
are saying "let me do this, beause I have at least some applications
where it is what I want -- yes, I know that other applicatoins may be
surprised at the behavior, but so what -- programmers should be careful."

As I have said many times, I have no objection to a specialized calendar
package that provides so-called Gregorian dates back to 4004 BC if that
is what you want, and indeed why not spend your effort writing such a
package!

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 (i.e. if the Pilgrims landed on a certain date, that date
should correspond to what the package deals with) is a truly bad idea!