Regarding [1], the move to Mesos was not a rewrite. Some core components had to be improved, but [2] was primarily an infrastructure upgrade that happened to ma

Regarding [1], the move to Mesos was not a rewrite. Some core components had to be improved, but [2] was primarily an infrastructure upgrade that happened to make Siri considerably faster by ditching hypervisors. It was a massive feat that was accomplished by relatively few engineers. There were always many components under review for reengineering; keep in mind that Siri was released before WebSockets was a thing, let alone RFC. The snowball effect of Siri's failures originated from iOS Program Management pushing us to expand into as many other languages as soon as possible from the beginning, thus diverting resources from implementing or polishing original "Siri, Inc." features and various "nice to have" improvements to en-US usage. Source: I worked on Siri from the moment it was acquired, and no longer work at Apple. Richard Williamson's recent comments are pathetic.