[erlang-questions] rpc is bad? (was Re: facebook chat server)

[Steve Vinoski]

On 5/23/08, Claes Wikstrom <klacke@REDACTED> wrote:
> Steve Vinoski wrote:
>
> > It's related to the fallacies, yes. The fundamental problem of RPC,
> > which is trying to make remote calls look local, is based on
> > essentially trying to ignore the problems covered by the fallacies or
> > trying to pretend they don't even exist.
>
>  I really like the above paragraph. Ages ago when I was naive, and I'd just
>  implemented distributed erlang with the basic concepts like
>  spawn/4, link/1 and most importantly {Name, Node} ! Message
>  I then set out to build higher level libraries on top of those
>  basic low level mechanisms. We still see the traces of those efforts
>  in the otp libs today.

Hey Klacke, all I can say is that I'm probably not alone in wishing
that ages ago I had been as naive as you claim to have been! :-) What
you've referred to as "basic" concepts in Distributed Erlang are IMO
amazingly powerful, succinct, and compelling.

>  Think rpc:multicall(), the pg and pg2 modules , mnesia hidden replication,
> etc.
>  Now some 10 years later I've never ever used those functions that try to
> hide
>  the network - even though I wrote them. Interesting.

Sometimes negative people will point out that my tastes in distributed
computing have changed over the years and that some of my recent
writings contradict what I wrote 10 or 15 years ago. My typical
response is "yes, I'm glad I can still learn!" ;-)

>  I have used rpc:call/4 extensively though but it doesn't hide the network.

Coincidentally, that's exactly one of the calls I had in mind when I
previously wrote that Erlang/OTP doesn't try to hide distribution from
the programmer.

--steve