the most interesting aspect of the article (to me at least) was the discrepancy about the perceived ease of locks and the actual rate of error while using them.
wonder if something similar would / could be observed while using message-passing (a la go / erlang) vs implicit or explicit locking...