[zeromq-dev] lacks

[Pieter Hintjens]

On Thu, Apr 24, 2014 at 10:46 PM, Andrew Hume <andrew at research.att.com> wrote:

> can you expand on this a little (or a lot)?
> One thing we lack in ZeroMQ generally is resilience, supervisor trees and
> such.

OK, off-the-cuff expansion:

It seems a missing piece in our puzzle; how to start message
processing tasks on demand, and handle their deaths smoothly. I'd
probably want to use messages to communicate such exits. I think we
could make this work using green threads, with internal or external
tasks that eventually plug in with some protocols.

I'd like e.g. to make a web server that uses such a mechanism to start
handlers for specific URI paths, track their presence, and where
handlers can cheaply die if they hit issues.

I'm no expert in supervisor trees, and haven't used Erlang. This is hypothesis.

-Pieter