RFC 3107, "Carrying Label Information in BGP-4", May 2001
Note: This RFC has been obsoleted by RFC 8277Source of RFC: mpls (rtg)
Errata ID: 2319
Reported By: Bruno Decraene
Date Reported: 2010-07-07
Rejected by: Adrian Farrel
Date Rejected: 2010-08-20
Section 5 says:
"A BGP speaker should not advertise this capability to another BGP speaker unless there is a Label Switched Path (LSP) between the two speakers."
It should say:
"An eBGP speaker should not advertise this capability to another eBGP speaker unless there is a Label Switched Path (LSP) or layer two interface between the two speakers." Or just remove completely that sentence.
An iBGP router should be able to set up an internal BGP session for AFI 1 / SAFI 4 toward a Route Reflector even if the Route Reflector is not capabble of forwarding MPLS packets (This case is even described in the section 2 of the RFC)
2) + layer two interface
If both router are connected by a direct (sub)interface, they should be able to exchange MPLS packets even if there is no LSP between them.
3) Remove the sentence
AFAIK, the point is now better addressed by draft-ietf-idr-bgp-bestpath-selection-criteria-01.txt
After discussions with the document author on the MPLS mailing list:
- The EBGP/IBGP distinction is not relevant here, as this does not properly
capture the notion of whether a BGP speaker is in the data path.
- The ability to send an MPLS packet from one router to another does not
necessarily depend on there being either a sequence of MPLS routers
between them or on there being an L2 connection between them. There might
be an L3 tunnel, for example.
Furthermore, we feel that no confusion has arisen as the result of the current text so that there is no harm in leaving it as it stands.