Re: [Idr] draft-ietf-idr-bgp-extended-messages-12 WG LC (5/24 to 6/7)

["Sriram, Kotikalapudi (Fed)" <[e-mail]>]

Robert,

>I listed 4 alternative solutions earlier in this thread already :). 
>Dropping NLRI no matter how rare it would be is not one of them.

I don't know what other BGP protocol enhancements would 
ever need to use extended message;
but so far we know that BGPsec "may" need to use it 
(though very rarely, if ever) -- see note below**.
And it seems clear that BGPsec would never need to drop the NLRI 
(including when the peer has not negotiated extended message capability).
This is because it is designed so that a BGPsec speaker can convert a BGPsec 
update with path signatures to an unsigned BGP message (for forwarding).
BGPsec speaker can/will certainly do this conversion under these 
two circumstances that matter for this discussion:
1. The BGPsec message (that is being forwarded) has a size larger than 4096,
and the peer has not negotiated extended message capability.
2.  The peer has negotiated BGP capability but not BGPsec capability.


**Note about BGPsec update size: A BGPsec message size 
can exceed 4096 octets if there are 40 or more unique ASes 
in the AS path (given that each AS's signature size is no more than 
72 octets with the ECDSA P-256 algorithm). 
The currently seen Maximum AS Path Length (eliminating prepends) is 15 
http://bgp.potaroo.net/as6447/ 

Sriram