recursive event routing algorithm

Re: recursive event routing algorithm

Hi
[color=blue][color=green]
> > i'm searching for algorithm able to detect and resolve conflicts
> > during recursive event propagation.[/color]
> Events using either recursion or message queues are often implemented
> as substitutes for full threading. Rather than re-inventing threading,
> or a subset of threading, why not simply use threading to handle events?[/color]

I should give the reason in my previous post - here it goes:

I have to implement such a framework as a project on my "Component Software"
course :) My main goal to gain some knowledge about implementing highly
reusable software, and implementing "strange" event propagation routines
is only the device...
I think my prof. choosen this subject, as implementation of such algorithm
probably would give lots of architecture design problems
(eg. how to make it "easy to use" from client point of view).

Hope make things clear...
Thank you for reply.

Marcin

Re: recursive event routing algorithm

zorba128@interi a.pl (Marcin Kielar) wrote in
news:ea0824bb.0 312141447.1e7cf [email protected] gle.com:
[color=blue]
> hello
>
> i'm searching for algorithm able to detect and resolve conflicts
> during recursive event propagation.
> I'd like to implement small message passing framework without using
> message queue - messages
> would be delivered by simple method calls. This leads to recursive
> message ordering instead of
> natural one - and thus sometimes may cause problems. I know there are
> algorithms that detect this
> situation and block problematic messages for later delivery. Such
> algorithm is used in blackbox
> (component pascal/oberon) framework, but i'm unable to find it's
> detailed description....
>
> any ideas???[/color]

Why exclude the use of message queues? They are frequently used with
good success.

A recursive approach provides the effects of a message queue using
process stack space. One problem with this approach is that many
OSs provide relatively limited stack space for a single process.

Events using either recursion or message queues are often implemented
as substitutes for full threading. Rather than re-inventing threading,
or a subset of threading, why not simply use threading to handle events?

The abstract model of an event handling system includes the idea that
events are asynchronous with regard to the responders, and that each
responder should be allowed to proceed independently of all other
responders. Such behavior has been implemented as a set of method
calls in a single thread, but at a cost of significant complexity.
This is one model that fits a multi-threaded design much better than a
single threaded design.

Jim Rogers