Joel Reymont's frustration with Erlang

There is a couple of problems with the current Erlang implementation that tends to nag me. One of the more problematic for newcomers is that the bytecode format doesn't contain line-number references from the original source. Hence, an error can not show you exactly where the problem occurs, but it can give you a backtrace. This is rather unfortunate for Erlangs adoption.

I wonder if this guy tuned his network card. OS X default for max open sockets is 128, can make it hard to max out the cpu running network based loads generators.<p>sudo sysctl -w kern.ipc.somaxconn=512<p><a href="http://www.macgeekery.com/tips/configuration/mac_os_x_network_tuning_guide_revisited" rel="nofollow">http://www.macgeekery.com/tips/configuration/mac_os_x_networ...</a>

Scaling something to 10K isn't easy no matter what system you use. Erlang isn't a silver bullet.

joel rules. he will give any useful tech a chance. he approaches problems like an everyman. he isn't afraid to ask for help. i've watched him churn through lisp, haskell and ocaml, and now erlang.<p>i think he should give haskell another chance. it clobbers erlang on every performance metric, including massive concurrency, which we can now say erlang no longer "owns". otherwise i think c++ is the only option

I'd be interested to know if his problems went away on a similar Linux setup.

And this is supposed to replace java? ;)