[techtalk] setting up a network.

[Snarfblat]

Deepa Karnad wrote:
> 
> Okay. So I first configure serial port. Then use ifconfig to set up link. I
> had read up an HOW-TO on Ethernet networking where there was a small part
> talking on null modem link. There it said, I first run getty from one box
> and check if link is up and working. Then I come out of getty and run pppd
> on both machines. ( I'm using PPP and I think the ppp modules are installed.
> Any way to check that?)
> This should explain why I enquired about getty. So, how are these two
> methods of obtaining a link different? In the end, all I want to do is route
> between 3 hosts using 3 links.
> 
> Deepa.

Ahh, ok.  If you will be using PPP and not SLIP, then yes, you
will need getty and
a few other things as well.

To clarify things a bit, I was under the impression you were
using SLIP.  PPP and PLIP
(PLIP == SLIP for parralell ports) are other options for
networking.

You would be better off using PPP, and I will continue from that
angle..
(i found this page: http://www.swcp.com/swcp-info/slip-ppp.html )

Ok, first off you will need getty only if you want to autheticate
users.  You probably dont want this.

Most of this info is in the HOWTO for PPP.. (
http://www.linuxdoc.org/HOWTO/PPP-HOWTO.html ) but its kinda of
confusing since you are using ppp for local networking.

Setup:
First you need to decide how you will connect the multiple boxes
together.  Because PPP requires
a client and server, you can either:

1) Make all clients connect to one central server, this would
route requests between clients and act as a gateway to any other
networks.  This is like an ISP setup for modem users.  The
downside of this is that you will quickly run out of serial
ports.

2) Run a client and server on each machine.  Somehow define which
client connects to which server, with the end result being all
are linked together in some fashion.  This would require two
serial ports for each machine.  Downside of this is the
complexity, there is a fair amount of work to do to configure the
machines correctly to make sure no machine is left hanging, and
also the path from one machine to another could be quite a few
hops, depending on how many workstations you have.

Now, regardless of which way you choose, you will have to then
configure the clients, and configure the servers.  Note that a
machine may be a client and a server depending on how you
implmenent the network.

For clients:
You will be using PPP over a null modem connection.  See
http://www.linuxdoc.org/HOWTO/PPP-HOWTO-27.html for how to do
this.. This section is explained fairly well.
The commands to pppd are what configure the network as far as IPs
and other network info is concerned.

For Server:
http://www.linuxdoc.org/HOWTO/PPP-HOWTO-26.html for specifics,
this section is also fairly good.
They also mention the kernel modules you will need.
You can ignore anything with standard dial up stuff.  You arent
going to have users dial in via modems  (are you?).

Hope that clears things up a bit.

-- 
.oO()Oo.oO()Oo.oO()Oo.oO()Oo.oO()Oo.oO()Oo.oO()Oo.oO()Oo.oO()Oo.
    coderman at mindspring.com | http://cubicmetercrystal.com/
         "You are the product of a mutational union 
           of ~640Mbytes of genetic information."