We are not in the dark ages anymore. Denying reality doesn't help anyone here. Shared memory is dead for concurrency, it can't be made neither fast, nor easy to use.
I don't know what reality I'm denying. These post-dark-ages days Amazon, Apple, Google, Netflix and Twitter -- among many, many others -- are betting heavily on OpenJDK, with its shared heap GCs, for many/most of their big, concurrent backend applications.
I though we were talking about engineering challenges, research, not bets execs of large companies make for very different reasons, mostly hiring reasons.
Oh, sorry, I didn't realize you're a Google/Apple/Netflix/Amazon executive who's intimately familiar with their technical decisions. It's certainly true there is a lot of research on non-shared memory, but also a lot of research on shared memory. If non-shared has so decisively won, then not only the executives but also the researchers didn't get that memo. If and when non-shared "wins", the popular industry platforms will adjust as well.
>I though we were talking about engineering challenges, research, not bets execs of large companies make for very different reasons, mostly hiring reasons.
Large companies don't pick technologies for hiring reasons, small companies do. Large tech companies can afford to train their engineers in whatever they use internally and have the money (and brand recognition) to not worry about needing technologies to entice candidates. They pick technologies to solve business problems.
