Why Events Are A Bad Idea (for high-concurrency servers)

I thought it was strange that in the '“Problems” with Threads' section they didn't even mention testability - which is the number 1 problem we have in trying to create high-reliability systems with threads.

I work in embedded systems which run unattended for months at a time so they have to be reliable. The interactions between threads can create all sorts of unexpected race conditions which are very hard to predict. And if you can't assure repeatable behavior you can't properly test a system. On the basis of testability alone we avoid threads.

Worth pointing out that one of the authors, Eric Brewer, is now part of the effort to design Google's next generation infrastructure:

http://twitter.com/#!/eric_brewer/status/68051541063503872

We just had this yesterday: http://news.ycombinator.com/item?id=2907415

He also tweeted this in relation to it:

http://twitter.com/#!/eric_brewer/status/105160325824917504

Its a response to http://twitter.com/#!/ericflo/status/104972704016379904.