Hey All!

I have an interesting problem that I am trying to work out regarding Fido
message clustering and would like to get some input from the knowledgeable
people here.

Basically what I have been setting up here is an array of machines,
actually just two at the moment, listening in on a wireless network.  One
of these machines acts as a hub to the outside world but I am considering
altering this soon so that if one hub is down then another takes it's place
and from the percpective of the outside world looks exactly the same.  From
the percpective of any machine on the local network, it of course will fall
back on it's internal list of hubs so that if the first one is unreachable
it will try the next one and so on if that one is down until it runs out of
hubs and "decides" it is the hub and contacts the uplink to the
outside world itself, appearing to that uplink exactly like the regular hub
who would normally take care of this issue.  To compilcate this a tad
further suppose now that there were two normal Fido hubs out there who are
unaware of the local network's ability, each having a unique Fido address,
and supplying the identical echoes to the local network, but only seeing
the address from the local network that pertains to it.  In essence if the
local machine cannot reach the first hub for whatever reason it could
fallback to the other outside hub but then would have to alter the headers
in all the outbound pkt's such that the outside hub would view them as
valid.  Additionally if that event were to occur then the local acting hub
would have to take this into account so that it wouldn't duplicate the
sending of these messages to the other outside hub as well as inform all
the other machines on the local network that these particular messages have
been dealt with so that they don't duplicate that particular event and can
all share in whatever messages that are new to them.  Internally this is
easier then remotelly (the uplinks) since standard Fido headers aren't
required but becomes complicated if the created pkt's headers have to
change to reflect the "new" address if the fallback has to occur.
 Right now I figure the simplest way is just to recreate the pkt's if the
preferred hub is unreachable but this could get to be a resource pig if it
were to happen too often.

Any suggestions?

Life is good,
Maurice

--- Msged/LNX 6.1.1