Re: Source address selection in IPv6 multihomed multi-addressed sites

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

Iljitsch;

> >    However, to enable source address filtering to discard packets with
> >    source addresses not belonging to an ISP, it is useful to enable a
> >    host, not some intelligent intermediate router, select a source
> >    address compatible with an outgoing ISP.  For that purpose, intra
> >    domain routing protocols or something like that should maintain
> >    routing table entries with not only preference values of an external
> >    routes, but also proper prefixes to be selected for source 
> > addresses,
> >    if the entries are chosen by a host.
> 
> > It should be noted that it is already doable with the current OSPF 
> > spec.
> 
> Hm, how would that work? In BGP you could see the next hop AS number 
> and map this to a source address,

And the information is carried by IGP as has been stated in my draft.

> but in OSPF there is no obvious way 
> to do this. (Although I'm sure a non-obvious way can be created.)

You should check the format of AS-external-LSAs, of, say, RFC2740,
where there is a place to hold addresses of outgoing routers.

> However, I certainly wouldn't want hosts to interact with OSPF as this 
> is a somewhat fragile protocol. In RIP you can simply ignore what hosts 
> have to say and in BGP you can filter it, but in OSPF as-is you can 
> only hope the host don't send any information that screws up the 
> routing table.

The paragraph above contains so much errors to worth commenting.

> And then there is still the problem of how useful this information is. 
> Even today with BGP is is fairly common that BGP selects a very bad 
> route

You can use whatever EGP you like, though BGP practically is the
only solution.

> What BGP doesn't know is that the 
> interconnect between B and C is 500 km away while the interconnect 
> between D and E is on another continent.

Relying on ASPATHLEN does not address the issue.

On the other hand, BGP administrators know that the interconnect
between B and C is 500 km away while the interconnect between D and
E is on another continent.

However, BGP administrators feel difficulty to grasp connections
between neighbor ASes, if there are so many neighbor ASes.

> Obviously this problem is only 
> going to get worse as the routing table becomes smaller. That's why I 
> think that there will always be upward pressure for the routing table 
> size.

The smaller routing table allows more policy based control,
not blindly relying on ASPATHLEN.

							Masataka Ohta