"Richard A. O'Keefe" <ok at atlas.otago.ac.nz> wrote:
> "Lex Spoon" <lex at cc.gatech.edu> wrote:
> 	I would really love to a see a system that smoothly integrates local and
> 	remote calls--I'm not attacking the idea in general.  It just seems
> 	that everyone makes the same first mistake of trying to make remote calls
> 	look *exactly* like local calls.
> 
> There is a language that does this:  Erlang.  In Erlang, sending a message
> to another process is <pid> ! <message>, no matter where the other process
> is located.  But isn't there a distinction between unreliable remote sends
> and reliable local sends?  No:  local message sends are unreliable too;
> they might in principle fail too.  Amongst other things, the process you
> are sending to might crash (it's not just machines that crash).  Of course,
> Erlang is a mostly functional language, and function call _is_ local and
> _is_ reliable in Erlang.
> 

So there's still a distinction between a "message" and a "function
call".

To be a practical programming system, it seems like you have to know
that some things are reliable.  If not, you couldn't even be sure about
"x bitAnd: 3".


-Lex