In programming lingo, communication by side-effects is when
two procedures communicate via side-effects. In other words
instead of doing:
procedure1(a,b)
procedure2(b,c)
to get c you do this:
procedure1(a)
procedure2(c)
Hey, where'd b go? That's what everyone who reads your
code will say. It is in there, but instead of being an argument, it
is now a global variable that is changed as a side-effect of
procedure1. So? So, no one can read your code for one, and
for another, you make you code less reusable since
procedure2 mysteriously depends on the state of some
variable(s) that are not arguments (where most people look first),
and may or may not be initialized and/or allocated properly except
when procedure1 works its magic.
If you see code like this, you have two choices. Make a new
procedure that does 1 then 2 and takes proper arguments, or modify
1 and 2 to communicate via arguments like normal people. Which you
want to do depends on whether or not you ever want to do 1 without
doing 2 or not.