[comp.protocols.tcp-ip] Reverse Path Forwarding

dnwcv@dcatla.UUCP (William C. VerSteeg) (01/20/89)

We are in the process of porting a proprietary PC networking 
system to an IP based system. One of the thing we would like to
do is use IP routers to forward broadcast messages to "all nodes"
of our network. We are defining "all nodes" to be "all systems
on my (subnetted) network". When a node broadcasts to the network address,
RFC922 states that routers should forward this broadcast to all
applicible networks. In other words, the broadcast 36.255.255.255
should reach all members of (subnetted) network 36, whether
they live on the same ethernet cable as the broadcaster or not.


I have noticed that some routers do not implement the 
algorithm (Reverse Path Forwarding) that allows this to happen.

Does anyone know of an implementation 
that does reverse path forwarding?


Bill VerSteeg
DCA
1000 Alderman Drive
Alpharetta Ga. 30201

P.S. To the vendor to which I have been talking about this problem-
     Notice how nice I was here and didn't shamelessly rake you over 
     the coals in public. Wasn't that nice?

hedrick@geneva.rutgers.edu (Charles Hedrick) (01/21/89)

As far as I know, cisco routers (the only ones I know in detail) do
not implement RFC922 (reverse path forwarding for broadcasts).  If
they did, I would disable it.  There are two many machines that still
don't know about subnets.  Thus it is very hard to know when a host
really wants a site-wide broadcast as compared to one on the local
subnet.  I also think 922 may be overly optimistic about preventing
loops.  But this is just a guess.  At any rate, I think if you want to
do a wide-area broadcast, you should be looking at IP multicasting,
not at network-wide broadcasts.  Multicasting is still viewed as
experimental, and is not widely implemented.  However I think you'd
stand a much greater chance of getting router vendors to implement
that than the RFC 922 scheme.  On the other hand, if I were doing a
product which I wanted to sell to people other than network wizards
(i.e. people whose network contains systems may be a year or two
behind the latest proposals), I'd not assume any kind of broadcasting
outside of the local physical network.  Also, beware of the various
odd configurations that people sometimes create, such as multiple
networks or subnets on a single cable.