Re: Software Engineering News Brief

[caip.rutgers.edu!halasz]

In article <55rmsc$2ee$1@shade.twinsun.com>, eggert@twinsun.com (Paul Eggert) writes
> Well, that's true as far as it goes, but it
> turns out that several Solaris applications (SCCS and xterm, for example)
> _do_ have year-2000 problems because the application developers saw fit
> to take the year modulo 100, or subtract 1900 from the year, or some
> such nonsense.

The latter is not nonsens, for if
	x < y
then
	x - 1900 < y - 1900
--there is nothing wrong with that.  But this
	printf("%02d", y - 1900);
amounts to year modulo 100 (if "y >= 0"), not year less 1900.  To wit,
by itself subtracting 1900 is no problem, but one who did is eazily led
astray into the real trap.

Incidentally, the century-turn is not enough grounds for to drop the
convenient abbreviation of uzing only the last twain digits of the
year-numeral; it is grounds for care.  In a system that I workd on, such
2-digit numerals were permitted in data-entry, in text-formd datafiles,
but the meaning of it depended on a configuration variable.  If this was,
o, 1940, 2-digit year-numerals span 1940-2039.  That is, if the short
numeral is less than the (0-based) remainder of dividing the variable by
100 it stands for the year numberd 2000 (100+variable-remainder) greater,
otherwize for the year numberd 1900 (variable-remainder) greater.

One can make SCCS work on by either taking the year in the file to be
1900 less than the real number--in this case, 2000 Jan 12 is "100/01/12",
or installing sum such scheme as aforesaid in "delta", "get", "prs", ...,
maibe make it a new flag in the history-file itself, & if not there it is
taken for 1970 (an important year for Unix).