"Sean Kirkpatrick" <Sean.Kirkpatrick at PipelineTrading.com> writes:
> Hopefully I'm not raising a flag for no reason here, but I believe
> I've found a bug within the generate_listen_paths subroutine in the
> init_cluster.pl script.? In the heart of the second set of loops,
> when a shorter path has been found, the following gets set:
>
> ??????? $VIA[$node3][$node1] = $node2;
> ??????? $COST[$node3][$node1] = $newcost;
>
> Shouldn't the reverse listen path also be set here?
>
> ??????? $VIA[$node1][$node3] = $node2;
> ??????? $COST[$node1][$node3] = $newcost;

I will take a peek; I expect this is not something requiring "much
ado," from two perspectives:

 1.  The reverse path would presumably be discovered eventually, so
     that this "shortcoming" should merely make the calculation run a
     bit longer, and

 2.  I'm hoping to make this code all go away anyways because of the
     following...

I'd like to eliminate the need to populate this at "initialization
time" in favour of a heuristic that would run each time that a
subscription is added.

Thus, when a node is initialized, listens would be set up that go to
and fro through whatever node appears to be "master" in the network.

When a subscription is activated on a node, all of the 'listens'
pointing at that node would be redirected to either point to it
directly, or to its 'parent.'

Dropping a subscription or a node would lead to pointing various nodes
to the parent.

It's just a matter of programming and testing...
-- 
"cbbrowne","@","ca.afilias.info"
<http://dev6.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)