IMHO trying to define "ad hoc" in terms of what it is is doomed to be flawed and it may lead to disagreements. For example, I don't really think that on average the set of ad hoc problems is easier (or harder) than the set of all problems.
I think that ad hoc problems have to be defined in the sense of what they are not. Ad hoc problem is a problem that can't be solved using known techniques and named algorithms only, it requires at least some problem-specific thinking. (Note that this doesn't exclude the application of known algorithms, the current text seems to be far more restrictive.)
--Misof 15:03, 28 Dec 2005 (EST)