RE: FW: 3gpp scenario 2

[Pekka Savola <pekkas@netcore.fi>]

On Thu, 19 Dec 2002, Karim El-Malki (EAB) wrote:
>  > Many run routing protocols on top of configured tunnels, 
>  > thus there is no 
>  > problem you describe.
> 
> Don't think this is something we should recommend.
> It would likely mean running a routing protocol to advertise the IPv4
> tunnel endpoint address and a v6 routing protocol within the tunnel to
> advertise the v6 routes. 

I don't see anything particularly bad about this; after all, it's a
temporary thing only (but if your underlying infrastructure is completely
backed up at the IPv4 layer, most wouldn't need to run v6 routing
protocols, of course).  People are doing much, much nastier things with
routing protocols (consider e.g. MPLS/BGP VPN's carrying IGP routes).

> If we can run something like MP-BGP over v4
> isn't that better? 

If that requires additional infrastructure, like upgrading your routers to
support the extension, configure full BGP mesh of peerings between all the
islands, or the like -- perhaps not.

But I'm not completely opposed to the idea: if, for some reason I couldn't 
fathom, the operator was running MPLS core network, MP-BGP over v4 just 
_might_ be a good solution if the number of the islands is high.

But I don't believe we should be designing to cope with every possible 
cornercase (whether this is one is another question) -- then we'd get 
nothing done, and would end up requiring all kinds of transition 
mechanisms -- I *don't* want, based on 3GPP analysis, that we end up 
requiring standardization of [BGP] or [IGP] because there's some 
perceived need for deployment.

-- 
Pekka Savola                 "Tell me of difficulties surmounted,
Netcore Oy                   not those you stumble over and fall"
Systems. Networks. Security.  -- Robert Jordan: A Crown of Swords