David Abrahams wrote:
IMO someone like you probably *should* vote against it, especially if you've taken the time to do reasonable review. At least you have a real-world need for FSMs. I just "know" that FSMs are sometimes used for lightweight speed-critical components. Or maybe more people should be voting for it on the *condition* that an option with fast dispatch must be available. I must say, in a world replete with high-performance generic components, I agree that "Boost FSM" is too broad to cover a library that only does dynamic double-dispatch.
As someone working in the game industry, I can say that right now, even if we use fully use Boost in our tools, in our engine we use Boost with parsimony (typically libraries .hpp without .lib; easier to control code size) and also disable RTTI since the memory overhead is not worth the cost. A performant state machine mechanism working without RTTI would be interesting. Regards, Nicolas