🚀 go-pugleaf

Ernest Clinton Arbaugh Jr (cea@cerc.wvu.wvnet.edu) wrote:
: 
: I've got PPP running on a pair of SPARCstations, but  I can't seem 
: to get them to route.   Here's basically what things look like:
: (Greatly simplified.)
: 
: Local Site    Remote Site 
: +----+        +----+
: |  A |--------|  D |
: +----+  PPP   +----+
:    |            |
:    |   <.   .>  |
:    |     \ /    |
: +----+    Y   +----+
: |  B |    |   |  E |
: +----+    |   +----+
:    |      |
:    |  <- Ethernet
:    |
: +----+
: |  C |
: +----+

When PPP is started up on A and D, somehow A must publish the route to 
the D/E network to hosts B and C, and D must publish the route to the A/B/C
network to E. (A and D, of course, have no problems with each other, but
have no knowledge of the other's second interface.) Routing protocols do
exist and can, I suppose, handle the situation, but I have had very little
luck dynamically adding routes when SLIP/PPP start up.

To get around this, we have added static routes for hosts on multiple
different networks to reach each other through the PPP/SLIP gateway hosts.
This works well for us.

If there is only a single remote host, the gateway host on the multi-host
network can respond to ARP's for the single remote if told to do so,
but that doesn't work very well when there are multiple remote machines.

An example gated.conf entry found on ALL our hosts on the main (25.4.0.0)
network is as follows:

    static {
	192.61.63 gateway 25.5.4.6;
	192.61.59 gateway 25.5.4.2;
	192.61.58 gateway 25.5.1.27;
	192.61.57 gateway 25.5.4.50;
	192.61.56 gateway 25.5.4.101;
	192.61.55 gateway 25.5.4.58;
	192.61.54 gateway 25.5.4.43;
	default gateway 25.5.1.1;		<== Cisco
    };

    propagate proto rip {
	proto static {
	    announce default;
	};
    };

Each line of static route lists the two interfaces on a host with a
(in this case SLIP, but PPP shouldn't differ) serial link. We subnet
the class C address so that the SLIP link itself takes two addresses
and most of the rest are available on the remote end Ethernet. In
some cases, the remote end consists of only one host and no Ethernet
and in others, there are a number of hosts and devices on the far end.

The gated protocol seems unable to deal with subnets so we simply send
everything to the gateway (on network 25.5.0.0) and allow them to route
appropriately.  (The gateway hosts also have static routes to direct all
192.61.xx.0 traffic to the remote end of the SLIP link, which then
forwards if necessary on its Ethernet interface.  Lots of routes, but
once they are set up, it all works flawlessly.

Regards,

Ted Rypma
HP Panacom Division
Waterloo, Ontario

(All opinions expressed are mine and not my employer's.)