I'm rather far behind in my mail, but hopefully current on this thread. A few comments on the adopted draft:
The thing I find most worthy in this document is the discussion about how excessive prepending increases your vulnerability to route hijacking. This issue will exist whenever you have a well connected peer that is willing to forge their AS_PATH. It's probably worth the general nod to bgpsec because it implies how you don't get out of that problem without locking down the contents of the as path. The well connected peer issue is inferred in 3.1. While I agree with Jakob's comments that lead to the updated section 3.4 text, trying to normalize "avoid other people's bugs" is awkward at best. 3.5 similarly fits in that category. Section 4 is a nice start to describing the solution space, but is problematic on a few levels: - Scoping communities from your directly attached ISP may not help you in some circumstances. - More specifics is a hard yank on traffic - if you're even permitted to originate them. If you're not lucky, your prefix is already so long from being a small service provider that you can't make it longer. - BGP origin code as a high order tie-breaker is a technique that works well, but only at places where as-paths are of equivalent length. For many situations resulting in long paths, I'm not sure that helps. Section 5's guidance of "no more than 5 prepends" is probably fine these days, but that is largely a matter of the current diameter of the Internet. Section 7 - Security Considerations - could likely use a paragraph discussing how long prepending may make you more vulernable to route hijacking. ---- Feedback on the contents done, a personal anecdote: Just prior to 2000, I used to work for a small regional ISP. That ISP was multi-homed to a tier one and two tier two providers, each of which were trying to fight their way into tier-one space. Our motivations for our multihoming were one part resiliency, one part cheapest bandwidth for the dollar, and one part the fact that connectivity at the time to specific sites meant that we had a strong reason to use a specific provider. It was our experience at that time that it was necessary to use prepends in the range of 3-5 for portions of our address space in order to try to provide appropriate inbound load balancing. The number was that high because at the points we were trying to balance traffic for, the denseness of the Internet topology meant that the unpadded routes were often getting closely tied for path length. Prepending wasn't a complete panacea. We regularly experienced traffic spikes after traffic rebalancing until we found a prepend length that was stable enough. I recognize at this point that we were suffering from the effects of BGP Wedgies. (See RFC 4264.) I've been out of the operator game for a very long time. We've generally seen the diameter of the Internet decrease since I was in the business. The problems I was solving at the time still exist, but aren't quite as severe in the well connected bits of the Internet. I would not be shocked if some of the far corners of the Internet still need this for similar motivations than my employer had twenty years ago. ---- I will offer one final bit of feedback: Some customers simply filter very long AS_PATHs. The issues noted in this draft are effectively self correcting in the presence of aggressive filter policies. On Tue, Sep 08, 2020 at 05:16:30PM +0000, Michael McBride wrote: > Hello wg, > > We have addressed all comments so far. Thank you. We still need to expand the > prepend alternatives section. Please give it another look and let's see what > else should be included. > > mike > > -----Original Message----- > From: GROW <[email protected]> On Behalf Of [email protected] > Sent: Tuesday, September 08, 2020 9:58 AM > To: [email protected] > Cc: [email protected] > Subject: [GROW] I-D Action: draft-ietf-grow-as-path-prepending-00.txt > > > A New Internet-Draft is available from the on-line Internet-Drafts > directories. > This draft is a work item of the Global Routing Operations WG of the IETF. > > Title : AS Path Prepending > Authors : Mike McBride > Doug Madory > Jeff Tantsura > Robert Raszuk > Hongwei Li > Filename : draft-ietf-grow-as-path-prepending-00.txt > Pages : 10 > Date : 2020-09-08 -- Jeff _______________________________________________ GROW mailing list [email protected] https://www.ietf.org/mailman/listinfo/grow
