[dmd-concurrency] draft 2

[Sean Kelly]

I'll simply add to Walter's, since I agree with his comments...

"In addition, message-passing APIs (such as the MPI specification [19]) have been avail- able in library form, initially for high-end hardware such as (super)computer clusters."

I understand that it's necessary to give a nod to message-passing, since it was present at the time as well.  But this sentence feels tacked onto a paragraph that's really about something else.  It might be preferable to drop this into its own paragraph with a few more sentences for context.  I believe the Actor model was first conceived in the 60s, Hoare came up with CSP in the 70s, and Occam and the Transputer were created based on this in the 80s.  Far too early for its time to succeed though, since the only machines with multiple processors were supercomputers (as you said).  Joe Armstrong said he based the Erlang design on CSP, but it many ways it feels more similar to the Actor model.  This is probably way more information than necessary and it may dilute the point, but I'd like to see at least a tiny bit of context for that sentence.

More once I've read draft 3.