The other stuff you mention are good for the time. Extra props for getting work done in direction of LISP machines. I could see that I'd have enjoyed using that system besides the odd quirks you mentioned. Plus, the name of the network is way cooler than the other guys'. ;)
re Magnus/ML. I had CakeML in my collection but didn't know about Jitawa + Milawa. I remember skipping it seeing "verified lisp" in abstract while rushing through ACM/IEEE stuff and thinking "I already have VLISP so who cares." However, I appreciate your link because Milawa was a great find: an answer to the "Do you trust the prover?" question that people keep asking. Plus, Jitawa verifying a LISP and it were a nice combo. Knew Magnus was smart but that's extra smart.
Anyway, looking at your update, I see CakeML now has a verified compiler and other cool stuff. I might stop recommending FLINT except for diversity. CakeML seems to be the one to build on with VeriML too limited imho: a portable compiler & runtime is always better than an interpreter for Getting Shit Done (TM). So, I appreciate that link.
Your HOL work might be tricky. I'll be interested in hearing your experience later as it will inform how other software/systems people might fare. If it's difficult or even before you start, you might want to check out Chlipala's methodology here:
It's more like the FLINT style than HOL. He's produced many solid works with that. His site (link on top) also has interesting tools for various use cases: Bedrock (verified assembly), Ur/Web (correct-by-construction web apps), contribute to Ynot (writing/verifying imperative code), and so on. What they built with it is obvious from descriptions. Their pace has been faster than full verification groups while knocking out lots of errors. There's also more investment of academic brains into Coq theorem prover so each work might benefit from others.
Now, a link to help you with what you're actually doing. The seL4 work at some point referenced this book on using Isabelle/HOL for verification. Might be what you need to learn it all. I'd try it first. Stay on it if you're able to do real stuff w/ projects you referenced. Otherwise, backtrack to try Chlipala's methods.
re browsers. Yes, they're critical for reasons you mentioned. There's a simple solution, though, which I used for years: physical separation w/ KVM switches and guards for information sharing. The browser stuff runs in a PC that handles it. Most of key work or trusted data on other system. The content is translated to formats that make verification easier or at least where you have trusted programs. The guard mediates flow of data to ensure files move back and forth without attacks on memory or transport. It's built as strong you can make it. KVM and drag-n-drop software make it easier to use. Tenix's Interactive Link is an implementation that's very similar.
re Worse is Better & Java. I semi-agree. Java is a horrid platform that was successful for these reasons: (a) similarity to existing, mainstream languages, (b) tooling in form of IDE's + libraries, (c) perceived advantages, and mainly (d) massive money in marketing by huge, enterprise firm. So, the strategy worked. Google, Mozilla, and Apple are all using the strategy with effectiveness. One needs deep pockets plus major connections to consumers and industry, though. Meanwhile, Python and Julia are examples (for me) of The Correct Thing with enough traits to make them successful despite lack of millions in funding, widespread industry support, etc. So, there's that approach too. Still an open topic where I have no definite answer.
re seL4. Good luck on the ARM, kernel, and HOL learning projects. Will keep you busy for some time I imagine with make a few less hairs on your head after righting with hardware and kernel-level issues. ;)
"Thanks a lot for this discussion and giving me hope WRT to intermediate ways to address the web browser problem!"
You're welcome and thanks to you, too. It's been an interesting one. Do give me feedback on the Concrete Semantics and Isabelle stuff later so I can assess minimal level of talent that can use it.
https://www.dreamsongs.com/Files/worse-is-worse.pdf
The other stuff you mention are good for the time. Extra props for getting work done in direction of LISP machines. I could see that I'd have enjoyed using that system besides the odd quirks you mentioned. Plus, the name of the network is way cooler than the other guys'. ;)
re Magnus/ML. I had CakeML in my collection but didn't know about Jitawa + Milawa. I remember skipping it seeing "verified lisp" in abstract while rushing through ACM/IEEE stuff and thinking "I already have VLISP so who cares." However, I appreciate your link because Milawa was a great find: an answer to the "Do you trust the prover?" question that people keep asking. Plus, Jitawa verifying a LISP and it were a nice combo. Knew Magnus was smart but that's extra smart.
Anyway, looking at your update, I see CakeML now has a verified compiler and other cool stuff. I might stop recommending FLINT except for diversity. CakeML seems to be the one to build on with VeriML too limited imho: a portable compiler & runtime is always better than an interpreter for Getting Shit Done (TM). So, I appreciate that link.
Your HOL work might be tricky. I'll be interested in hearing your experience later as it will inform how other software/systems people might fare. If it's difficult or even before you start, you might want to check out Chlipala's methodology here:
http://adam.chlipala.net/cpdt/
It's more like the FLINT style than HOL. He's produced many solid works with that. His site (link on top) also has interesting tools for various use cases: Bedrock (verified assembly), Ur/Web (correct-by-construction web apps), contribute to Ynot (writing/verifying imperative code), and so on. What they built with it is obvious from descriptions. Their pace has been faster than full verification groups while knocking out lots of errors. There's also more investment of academic brains into Coq theorem prover so each work might benefit from others.
Now, a link to help you with what you're actually doing. The seL4 work at some point referenced this book on using Isabelle/HOL for verification. Might be what you need to learn it all. I'd try it first. Stay on it if you're able to do real stuff w/ projects you referenced. Otherwise, backtrack to try Chlipala's methods.
http://concrete-semantics.org/
re browsers. Yes, they're critical for reasons you mentioned. There's a simple solution, though, which I used for years: physical separation w/ KVM switches and guards for information sharing. The browser stuff runs in a PC that handles it. Most of key work or trusted data on other system. The content is translated to formats that make verification easier or at least where you have trusted programs. The guard mediates flow of data to ensure files move back and forth without attacks on memory or transport. It's built as strong you can make it. KVM and drag-n-drop software make it easier to use. Tenix's Interactive Link is an implementation that's very similar.
re Worse is Better & Java. I semi-agree. Java is a horrid platform that was successful for these reasons: (a) similarity to existing, mainstream languages, (b) tooling in form of IDE's + libraries, (c) perceived advantages, and mainly (d) massive money in marketing by huge, enterprise firm. So, the strategy worked. Google, Mozilla, and Apple are all using the strategy with effectiveness. One needs deep pockets plus major connections to consumers and industry, though. Meanwhile, Python and Julia are examples (for me) of The Correct Thing with enough traits to make them successful despite lack of millions in funding, widespread industry support, etc. So, there's that approach too. Still an open topic where I have no definite answer.
re seL4. Good luck on the ARM, kernel, and HOL learning projects. Will keep you busy for some time I imagine with make a few less hairs on your head after righting with hardware and kernel-level issues. ;)
"Thanks a lot for this discussion and giving me hope WRT to intermediate ways to address the web browser problem!"
You're welcome and thanks to you, too. It's been an interesting one. Do give me feedback on the Concrete Semantics and Isabelle stuff later so I can assess minimal level of talent that can use it.