[Idr] Interpretation of multiprotocol extensions for IPv4

[Ondrej Zajicek <santiago at crfreenet.org>]

Hello

We found that some widespread BGP-4 implementations (Cisco and Quagga)
reject BGP session when the other side sends OPEN message that contains
capability advertisement (RFC 3392) without multiprotocol extension
capability (RFC 2858). When multiprotocol extension capability with
IPv4 AFI is included, the session established and IPv4 prefixes are
exchanged, but using traditional way (does not use MP_REACH_NLRI).

That leads to question about interpretation of multiprotocol extension
capability with IPv4 AFI (MEP_IPv4). I can imagine two possible
interpretations:

1) MEP_IPv4 should be advertised iff router accepts IPv4 prefixes in
MP_REACH_NLRI and MP_UNREACH_NLRI extended attributes.

2) MEP_IPv4 should be advertised when router accepts IPv4 prefixes in
traditional NLRI.

The first interpretation is IMHO what RFC 2858 currently says.
If this is correct then Cisco and Quagga behavior is
unreasonable.

The second interpretation is consistent with Cisco and Quagga
behavior and might be more useful (as it allows to specify that
router does not accept traditional NLRI, for example IPv6 only
router).

What interpretation should be considered correct?

-- 
Ondrej 'SanTiago' Zajicek (email: santiago at crfreenet.org)
OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net)
"To err is human -- to blame it on a computer is even more so."