News for hackers. Stuff that matters.
Re:Asynchronous IO in Rust by dboreham on Wednesday September 16, 2015 @11:46PM (#10231468)
No, backwards: the C10k techniques are the solution to the problem "how do I achieve my performance and scaling goals given the characteristics of the OS, languages, tools available to me today?". The current thread (sic) is asking the question "if we can change the language, tools (perhaps the OS too), what's the best approach?".

And: I have to counterbalance the assertion that Node.js is in any way good for anything besides "I need to code in JS, but not in the browser". I'd rather use VAX assembler and the $QIO syscall, typing on a VT100.

Re:Asynchronous IO in Rust by cwp on Thursday September 17, 2015 @02:46AM (#10231919)
That's exactly the point. So my question remains: if threads are the right answer today, either the C10K folks were wrong or something has changed since then. What?
Re:Asynchronous IO in Rust by dboreham on Thursday September 17, 2015 @09:23AM (#10233170)
The question is a different question, hence the answer is different. C10k wasn't considering the option of new languages, this thread is. The same question could have been asked back then (although most folks wouldn't have considered using anything other than C/C++ for high-scale servers then). Perhaps the thing that's changed is more memory and CPU power has made the option of new (and less efficient than C) languages a practical choice for new projects. I know that if, back then, I'd proposed the solution to my product's scaling issues with network I/O to be : "re-write all 500k loc in a different language", I'd have been laughed at.
Re:Asynchronous IO in Rust by krenoten on Thursday September 17, 2015 @03:28AM (#10232053)
Modern operating systems are better at handling many threads.