frame relay with ipx split horizon

[fujin.huang]

Hi,

On the cisco doc http://www.cisco.com/warp/public/116/fr_faq.html#21 it says "Certain protocols such as AppleTalk, transparent bridging, and IPX cannot be supported on partially meshed networks because they require split horizon"

but there is a command "no ipx split-horizon eigrp 100", what is this command for?

[MickPhelps]

IPX doesn't require split horizon, IPX RIP does. NLSP or EIGRP could be used in place of IPX RIP over partially meshed Frame Relay.

The article says "...split horizon (a packet received on an interface cannot be transmitted over the same interface...)"

This isn't split horizon. Split horizon is "a route learned from an interface won't be advertised back out that interface".

If there is some other issue of packets not being sent out the same interface they entered, I'm not aware of it. Anybody know differently on that?

Mick.

[fujin.huang]

So, in this case, if I implement "no ipx split-horizon eigrp", this will only ensure your routing table being correct, which means you will have all the routes even if you network is partially meshed.

But the correct routing table does not ensure the packet being forwarded properly, when a packet received on a hub router multipoint interface, it will not be forwarded to other spoke router. Because of this, ipx doesn't allow you to implement a multipoint interface with partial mesh. Am I correct to say that?

Anybody got any idea?

Thanks. fujin

[MickPhelps]

Your question is correct, but I'm sure it would work just fine... why else could you disable split horizon for IPX EIGRP if it wouldn't move traffic properly.

Mick.

[svermill]

I would think that the risk of not routing packets would exist if you didn't disable split horizon (how would spoke A learn of spoke B, etc?). But I guess in this case you just have to live with the slight risk of a routing loop every now and then. But count to infinity would deal with that I suspect.

But isn't the whole premise a little suspicious. Is there something preventing the use of p-t-p subinterfaces?