[1] www.amazon.com/dp/0486428753
There are indeed precise conditions that guarantee that a function can (a) be integrated and (b) the integral can be differentiated, and there are subtly different definitions of "integral" that lead to different sets of extremely weird functions being integrable or not. I can't remember any of the details despite having come across them many times.
In software culture, it seems much easier to talk in broad strokes when appropriate and in precise details when appropriate. You might get an argument, but it's usually about something relevant.
That urge is stronger in mathematicians than in software engineers.
Yes he is. The real solution is to banish the term "indefinite integral" from the language, and substitute "antiderivative". We should also not use the integral sign for antiderivatives, but unfortunately there is no well known alternative. :-(
Using the same terminology and symbols for antiderivatives as for (definite) integrals is confusing, as here. It also makes the Fundamental Theorem of Calculus -- a very profound fact if there ever was one -- appear to be something trivial about getting rid of limits on an integral sign. But consider: one can compute a definite integral using an antiderivative ... whodathunkit?
I had an abstract algebra professor, in my first year of grad school in math, who referred to operators D that satisfy the product rule D(fg) = D(f) g + f D(g) , which are not necessarily defined in terms of a limit, as "derivations". This is a natural generalization of the derivative, and apparently "derivation" is the correct technical term, but for a week or so I thought he was just not speaking English properly.
I'm not an algebraist now.