Re: and then

[mcsun!sun4nl!cwi.nl!dik@uunet.uu.net (Dik T. Winter)]

In article <1993Apr10.010355.4244@nosc.mil> sampson@nosc.mil (Charles H. Sampso
n) writes:
 >      In the original requirements for Ada (Steelman), functions were not
 > allowed to have side effects.  At least one of the candidate languages,
 > Blue if I remember correctly, took a stab at prohibiting them with rules
 > about exactly what kind of code a function could and could not contain. 
 > Somewhere along the way the requirement vanished, either by explicit action
 > or benign neglect, probably because the requirement was considered draconi-
 > an.  By that time I was pretty far out of the loop.

The requirements vanished gradually.  At one point there was not a dichotomy
but a trichotomy: functions (pure, no side effects), procedures (pure) and
value-returning procedures.  For optimization purposes this is the best, but
value-returning procedures fell out and functions with side-effects came in.

The basic reason for the shift to value-returning procedures was that it is
impossible to write such a basic thing as a random number generator without
side-effects.
-- 
dik t. winter, cwi, kruislaan 413, 1098 sj  amsterdam, nederland
home: bovenover 215, 1025 jn  amsterdam, nederland; e-mail: dik@cwi.nl