Conversation

Replying to
It's fine to experiment and build a cost-optimized elevator that can handle exactly its listed weight before crashing down, might even be fine to use for specialized applications. Software however tends to install these risky elevators in every building (but optimized for perf)
2
1
Replying to and
Sure, danger of sharp knife/memory-unsafe code decreases as code base matures, but... Google's Chrome team aren't there yet twitter.com/newsycombinato Neither is SSH twitter.com/TheHackersNews Even when safeguards catch problems, it's easy to mess up the fix
Quote Tweet
This was interesting, a minor bug report from a fuzzer was fixed incorrectly leading to a far more serious bug that the fuzzer never found. I don't know what the lesson is, but ¯\_(ツ)_/¯ /cc @hanno twitter.com/ProjectZeroBug…
1
2
Replying to
There will always be bugs in software and hardware. This being said, our civilization is litterally built on C/C++ systems (Linux, Windows, Databases, iOS, Android) from people who deal with very sharp knives. Much of our crypto infrastructure is written in assembly.
2
1
Replying to
That infrastructure will always collapse and that much of it is old is no excuse for not following modern construction norms, question is if SW can self-regulate or regulators need to step in. Some low-level primitives will always exist, asm in crypto to prevent timing attacks...
2
1
Assuming one language does take the throne from C/C++, even the ultimate winner may have a significantly smaller niche than C/C++ ever had - for a large part of their life they were used as general purpose languages rather than just "system" or "perf matters" languages.
2
1
Agreed. Hopefully we'll be left with a systems language focused on system stuff. My bet is on Rust or a future entrant. Swift is an Apple thing and Apple things usually stay Apple things. Of course, if Apple takes over the world, we'll all be writing OSes in Swift I guess.
3
1
Whether or not Rust displaces C substantially, it making these features usable enough to have widespread adoption makes it very successful. It's still pushing the boundaries of the region typing model, including very nice libraries designed around it, safe data parallelism, etc.
2
1