3 ms·
SPARK is the only one still going strong in verifiable, systems, programming space. I had hopes for ATS and BitC. There's Rust tie-ins but it's not a verifiable
by nickpsecurity 10y ago
SPARK is the only one still going strong in verifiable, systems, programming space. I had hopes for ATS and BitC. There's Rust tie-ins but it's not a verifiable language. There is ongoing work to produce necessary models to make it one but it's quite complex. It's fine, though, as languages like it, Go, and Wirth's were really about raising the baseline safety/security of average code. Way better goal. Whereas, the verifiable languages are about making the highest level of assurance easier even if one must sacrifice some benefits of popular, complex languages.
So, we have SPARK. After much work, C can be subsetted with HOL and AutoCorres. Not really designed for it, though. So, we just have one proper for imperative with another made to work with monumental effort.
Far as functional, we have quite a few options that are prototypes with some production use. Haskell & HOL used with seL4 plus plenty others. Tolmach et al did Habit system language based on Haskell but I think it's on pause. ATS seemed quite practical but hadn't heard anything on it recently. COGENT is a new one for verified, functional, systems language that compiles to C with IIRC certified translation. The ML's have many practical implementations and were designed for theorem proving. CakeML latest in that. PreScheme was verified with VLISP and could be rebooted. There's also model-driven and DSL approach from Hinjtens at iMatix, CertiKOS with DeepSpec team, and Ivory at Galois.
So, this is what we're at since BitC folded. Anyone know of any work I missed with at least one deployment of low-level, OS-style code with realistic, verification capability?
- glibgil 10y agohttp://www.sparkel.org/ http://www.sparkel.org/
- nickpsecurity 10y agoSPARK for ParaSail? Hell yeah! Didn't know about it. Thanks for the link.