[prev in list] [next in list] [prev in thread] [next in thread] 

List:       boost
Subject:    Re: [boost] [future] Early draft of wait for multiple futures
From:       Frank Mori Hess <fmhess () speakeasy ! net>
Date:       2008-05-31 0:37:35
Message-ID: 200805302037.35981.fmhess () speakeasy ! net
[Download RAW message or body]

[Attachment #2 (multipart/signed)]


On Friday 30 May 2008 20:21, Frank Mori Hess wrote:
> On Friday 30 May 2008 17:41, Giovanni Piero Deretta wrote:
> > Hum, I think that boost already has such a dispatcher: it is called
> > asio::io_service ;).
> >
> > IMHO futures wait sets should be used to wait for very few events. If
> > the number of events grows significantly, it is probably a sign that
> > the application would be better redesigned as an event driven state
> > machine. I do not think it is worth optimizing futures for large N.
>
> That's essentially where I started.  To do an event driven design
> requires the ability to generate an event when a future completes.  That
> means a completion callback as part of the future's public interface.  A
> completion callback has been resisted as too dangerous, and waits on
> groups of futures offered as a safer alternative.  Therefore, I'm trying
> to insist the wait functionality be sufficient to build a decent
> scheduler on.

What if we specifically support asio for handling future completion events?  
That is, the future lets you register an asio::io_service along with a 
completion handler, and the promise just posts the completion handler to 
the io_service when it is fulfilled.  Then the completion handler won't 
actually get run in the promise-fulfilling thread.  Would that be safe 
enough and is it sensible (I don't know asio very well)?

[Attachment #5 (application/pgp-signature)]

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[prev in list] [next in list] [prev in thread] [next in thread]