[davidtgoldblatt]

Hah, I think I'm that person you're referring to. I would characterize my response differently.

Stripping out the specifics, the problem was: "write some code that blocks until one of a set of memory locations changes its value". The Haskell/STM API provides exactly that interface (where the memory locations you block on are those that you looked at in the transaction). Rephrasing then, the question was "how would you implement a simplified version of the Haskell STM blocking mechanism?", and your solution was "I would use the Haskell STM blocking mechanism", answering the question as though it were one of API usage rather than of API implementation.

That's probably fine as a question of workaday engineering (and maybe an argument for Haskell over C++ on the basis that it has more extensive libraries built-in), but somebody's gotta write that STM implementation, and that's what the challenge was trying to get at.

(In fact, the GHC STM implementation does exactly what the solution I outlined does: with each TVar, it keeps a pointer to threads parked waiting for the TVar's value to change, and unparks all those threads when a transaction modifying the TVar commits).

[oconnore]

Fair. Sorry for misrepresenting you.