[Spread-users] Re: Please help me understand

apocalypznow apocalypznow at gmail.com
Wed Feb 9 16:08:21 EST 2005

Thanks Mike for drawing the diagram.  Your diagram seems to reflect what 
I said.... but because the diagram is distorted in my news reader, I am 
not 100% sure of it.  Here is another description, just for clarity: is a public ip, has a router, from which and are behind that. is another public ip (on 
another network), has a router, from which and are behind.

The answer (to Q #2) that John Schultz gave implies to me that it can be 
done, but you've said otherwise.  Why is that?

Mike Perik wrote:
> I think what apocalypznow is saying is that 
> and are the public ip addresses that two routes have and behind one router are two machines and and behind another router are two machines and
>         ____________                                    ____________
>        |   |--------------------------| |
>        |____________|                                  |____________|
>           /                     \                                           /                      \ 
> I'm not sure what the spread.conf would look like in this situation.
> I think the answer that John gave to question # 2 implies that no you can't do this.
> Correct?
> Mike 
> On Wednesday 09 February 2005 01:56 pm, John Schultz wrote:
>>apocalypznow wrote:
>>>Hi I'm a noob... so please bear with me.
>>>I am trying to set up 4 machines on 2 networks, which would all join the
>>>same group "chat01" and hopefully be able to message each other using
>>>multicast() with either RELIABLE_MESS or FIFO_MESS.
>>>Two machines are at, and their ips are and
>>>  Two other machines are at and their ips are
>>>at and  Port 4803 is open on each of the
>>>two routers.  Networks and are only connected
>>>on the internet.
>>>I've tried various configurations in my spread.conf file and I can't get
>>>all 4 machines to see any messages sent to the same group "chat01".
>>>1) How should my spread.conf be set up - specifically the segments and
>>>anything else?
>>You should probably have two segments corresponding to the two separate
>>networks -- assuming that by "networks" you mean that machines within

>>these "networks" can 1 hop broadcast or multicast to each another.
>>>2) Where do I run the daemons?
>>You should run the daemons on machines whose IP addresses are accessible
>>  (i.e. - routable) by all other daemons in the config file.  So the
>>daemons should be run on machines in those networks that can reach the
>>daemons on the other network.
>>>3) How many daemons should I run?
>>This is up to you.  But generally you should run at least one daemon per
>>network that will have clients and usually even run 1 daemon on each
>>machine that will have clients.  You run more daemons both for network
>>efficiency and for fault tolerance.
>>>4) To which daemon should each machine connect to?
>>Each machine should connect to the closest daemon that it can.  Usually,
>>   this is often a connection to a daemon on the same local machine.
>>However, since it seems your client machines will be on private networks
>>that would not be routable by all daemons, you will probably need to
>>remotely connect to the closest daemon with a public IP.

More information about the Spread-users mailing list