3 ms·
> b) people are actually ready to shell out some decent money for getting the verification done. That's true, but may be changing. > a) the state space is muc
by pron 9y ago
> b) people are actually ready to shell out some decent money for getting the verification done.
That's true, but may be changing.
> a) the state space is much smaller than for software (although it can still be prohibitively large for complex designs and brute force formal methods)
That's true as well, but that is why some formal methods -- like my favorite, TLA+, and all the sound static analysis tools (they work by abstract interpretation) -- try to tackle the problem of software verification by abstraction (in a precise mathematical sense, namely a description of the system using a sound approximation of one with fewer states). In TLA+ you can even describe your system in different levels of detail and show that they form a sound abstraction/refinement relation. This is not to say that it's a silver bullet -- it takes work and thought, and isn't appropriate for everything, but a system like TLA+ lets you at least choose the level of detail you wish to work in and so the cost you'll have to pay. The lower the level, the more confidence you get, but more work you need to do. The early signs of TLA+'s success in industry show the practice to be worthwhile in some/many very important -- and quite common -- circumstances.