Re: [sidr] IXP and Route Server and Next Hop transparency

[Robert Raszuk <[e-mail]>]

Hi Sandy,

> The owner of the advertised prefix doesn't have ownership of the next
> hops along the path. nor does it have anything to say about the
> legitimacy of next hops along the path.
>
> There's this whole "third part next hop" concept which would make
> determining legitimacy of the next hop complicated.

Third party next hop (typical in IX cases) assures that the ownership of 
the next hop stays unchanged. So I would observe that this would make it 
actually easier there.

Moreover there are deployed applications which specifically mandate to 
not change next hop across AS boundaries. As example I could bring 
Inter-AS option C for L3VPNs. Note that there can be transit ASes in the 
path too. So how are we going to assure a customer of such service that 
the advertising prefix from customer site attached to AS 1, transiting 
via AS 2 and terminating at AS 3 got to the final site on the other side 
uncompromized ?

Maybe one should ask the bigger question: Is BGPSec as is being defined 
in SIDR applicable to other address families other then 1/1|2 and 2/1|2 
which still carry IP prefixes or is it out of scope ?

How about Internet as a VPN (aka Internet in a VRF scenarios) where 
internet routes may travel as vpnv4/vpnv6 updates ?

How about those address families in BGP which are designed to carry 
control plane information between domains for example: bgp flowspec or 
rt-constrain ?

I am just trying to understand if we are talking more of enhancements to 
bgp security (bgp being the protocol) or perhaps about just limiting the 
SIDR scope to build a new layer for controlling traditional Internet 
prefix propagation ?

Many thx.
R.