*sigh* let's not go overboard on what can and can't be called a "context", either.
So-called "Boolean context" is a useful concept, especially if you are dealing with overloaded objects. I agree that it is important to realize that it isn't a "context" in the same way that "scalar context", "list context", and "void context" are and that using something in a "Boolean context" is hard to distinguish from using it in a scalar context (though it can be done by using overloading). I'd love to see an enhancement that elevates "Boolean context" to being a "first class context" (well, at least "second class" if you don't count "void" as "first class") by, for example, adding a OPf_WANT_BOOL flag.
"Interpolative context" can also be a valuable concept for explaining why these are different:
even though it is very much different than the "big three" contexts. This one reminds me of the "numeric", "string", and "integer" pseudo contexts.
For built-in Perl operators, [void context] is identical to scalar context
Doing a quick search, I found that keys knows the difference between scalar and void context so that it can avoid some work in the latter case. I suspect there are more than a few other cases. I'm not trying to contradict you, I just don't want others to overinterpret what you are saying.
I guess that by "do something different" (in the case of a user-defined function detecting void context) you are referring to "side effects" and that you are asserting that no built-in functions/operators of Perl have different side effects between scalar and void context. I would certainly be somewhat surprised to find a case that violates such an assertion but I wouldn't feel comfortable making that assertion myself.
I'd take issue with your using "context" to describe the special way defined works but that would just seem ornery of me. A similar "context" that I'm finding increasingly interesting is the "pseudo code block context" that is given to the first argument of map or grep:
- tye (but my friends call me "Tye")