Jeffrey Haas wrote:
draft-bonaventure-bgp-redistribution-01
This draft included: IDRP style DIST_LIST_INCL, DIST_LIST_EXCL Proxied NO_EXPORT Proxied Prepending
The IDRP-style DIST_LISTs seem to generate most of the heat. We never got a firm feel for why the other two componenents were disliked.
Geoff Huston proposed draft-huston-nopeer-00.txt to attempt to address some of the route propagation issues that the DIST_LISTs were intended to address.
The problem, as I saw it, was that in attempting to specify a subset of the routing space the authors specified an enumerated list of AS's that formed the boundary of this subset. The two major problems, as I see it, is that you may not have up to date information about what As's are on the boundary of the subset which you want to apply to the redistribution, and each remote AS that is not on the boundary has no knowledge whether it is intended to be inside or outside this set. The alternative approach was to specify a common condition which characterized all members of the subset, allowing each remote AS to use local knowledge to see if it met the originator-defined constraints or not. This was the basis of the no-peer approach. Geoff Huston Telstra