[OpenSIPS-Devel] RFC: new opensips design

[Dan Pascu]

On Friday 07 November 2008, Iñaki Baz Castillo wrote:
> 2008/11/7 Bogdan-Andrei Iancu <bogdan at voice-system.ro>:
> > Why? - I made a summary here -
> > http://www.opensips.org/pmwiki.php?n=Development.NewDesign#toc1
>
> Hi, I read:
>
> ----------
>  2.4  Asynchronous processing
>     * entire processing in core must be based on a asynchronous
> reactor - there will
>       be dead-time/idle because of I/O ops. The processing capacity
> will dramatically
>       increase.
> ----------
>
> AFAIK Reactor Pattern also waits in I/O opperations. I've read about
> Proactor Pattern that doesn't wait for them:
>   http://www.cs.wustl.edu/~schmidt/PDF/reactor-siemens.pdf
>   http://www.cs.wustl.edu/~schmidt/PDF/proactor.pdf
>
> Well, I have no idea about both of them XD

Terminology aside, (nowadays everybody comes up with a new term for the 
same thing on the argument that their implementation does it slightly 
different), the point of the reactor pattern is to react on events when 
they happen and to switch context back to the reactor when an operation 
would block, allowing the reactor to chose another readily available 
event to process. I'd say to avoid confusion with too many terms we 
should use the reactor term to have the meaning above, as this is how 
it's implemented in all the cases I'm aware of (I know of no reactor 
implementation that waits on I/O).

-- 
Dan