As usual, discussions have gotten off on various tangents. My proposal is to replace the safe_bool idiom used in boost::tribool with a simple operator bool. I maintain that this will a) not break anything already working. b) make tribool better model tri-valued logic c) address a failing in gcc which inhibits the current tribool from being constexpr. d) So that all in all we will be in a slightly better place. That's the whole thing. It's not about a) whether or not tribool itself is good idea b) whether tribool is correctly named c) whether implicit conversions to bool in other cases like optional are a good idea d) or anything else other than the above. My intent to to get this PR accepted so I don't have to clone tribool into my safe numerics library. Selfish I know - but there it is. So far no one has posted any reason why my proposal would not be an improvement. Robert Ramey