AS_PATH length and route selection (WAS: Strange BGP announcement)
At 06:03 PM 11/17/98 -0500, Vijay Gill wrote:
You appear to have misunderstood the problem. The issue has nothing to do with route selection, or are you talking about FastPath(TM, Pat Pending) here?
Cute, Vijay. Really cute. Of course, if everyone had only licenced it, we wouldn't have this problem now, would we? :p
The issue is that an invalid as-path attribute was injected from somewhere. The cisco as-path sanity check code failed to pick it up.
[SNIP] I have not misunderstood the problem, you have misunderstood my response. Someone posted that the routers should follow the RFC. I was commenting that you should be careful what you wish for. If we magically made all routers strictly conform to the RFC instantly, it would have solved this particular problem, but it would have caused *much* larger problems for the 'Net at large. I was also hoping to start a debate by those more knowledgeable than I about the ... wisdom of intentionally removing what is currently far and away the most used metric for route selection from the next version of BGP; namely: AS_PATH length. (At least I was under the impression that AS_PATH length was the used more than all other metrics combined for route selection in the backbone.) RFC1771 does not list AS_PATH length as a selection criteria - but at least RFC1771 does not specifically forbid use of AS_PATH length, as the draft for the next version does. Or perhaps I'm just misinformed. I have not actually read the draft for BGP 4+ (or whatever they're calling it now). This is second hand information, but I consider it to be from a very reliable source (even if he doesn't like ciscos ;). I am also not saying removal of this metric is a gigantic mistake, but my thought process to date leads me to tentatively believe it would be ... suboptimal. However, the people writing the spec obviously know more about it than I do, perhaps I've overlooked something. So, would anyone care to discuss (from an operational POV) what removing AS_PATH length from the BGP route selection algorithm will do to the Internet?
/vijay
TTFN, patrick P.S. Before anyone goes off on me, I once again *agree* that the least cisco could do is deal with malformed announcements and the errors they generate properly. There are just other parts of the RFC with which I'm not so sure I agree. I Am Not An Isp www.ianai.net "Think of it as evolution in action." - Niven & Pournelle
On Tue, 17 Nov 1998, I Am Not An Isp wrote:
Cute, Vijay. Really cute.
Sorry. I tried. God knows, how I tried, but alas, the spirit was willing but the flesh, it was weak.
Of course, if everyone had only licenced it, we wouldn't have this problem now, would we? :p
If we had a working draft to take a look at it, other than the marketspeak on the web page, yeah...... And if we had some cheese, we could make a ham and cheese sandwich, if we had some ham.
I have not misunderstood the problem, you have misunderstood my response. Someone posted that the routers should follow the RFC. I was commenting that you should be careful what you wish for. If we magically made all routers strictly conform to the RFC instantly, it would have solved this particular problem, but it would have caused *much* larger problems for the 'Net at large.
How so?
I was also hoping to start a debate by those more knowledgeable than I about the ... wisdom of intentionally removing what is currently far and away the most used metric for route selection from the next version of BGP; namely: AS_PATH length. (At least I was under the impression that AS_PATH length was the used more than all other metrics combined for route selection in the backbone.) RFC1771 does not list AS_PATH length as a selection criteria - but at least RFC1771 does not specifically forbid use of AS_PATH length, as the draft for the next version does.
Exactly. The RFC says the following The selection process is formalized by defining a function that takes the attribute of a given route as an argument and returns a non- negative integer denoting the degree of preference for the route. The function that calculates the degree of preference for a given route shall not use as its inputs any of the following: the existence of other routes, the non-existence of other routes, or the path attributes of other routes. Route selection then consists of individual application of the degree of preference function to each feasible route, followed by the choice of the one with the highest degree of preference. So, if a router conforms magically to the RFC, no big issue, it is left up to the implementors to decide how to do the selection.
Or perhaps I'm just misinformed. I have not actually read the draft for BGP 4+ (or whatever they're calling it now). This is second hand information, but I consider it to be from a very reliable source (even if he doesn't like ciscos ;). I am also not saying removal of this metric is a gigantic mistake, but my thought process to date leads me to tentatively believe it would be ... suboptimal. However, the people writing the spec obviously know more about it than I do, perhaps I've overlooked something.
So, would anyone care to discuss (from an operational POV) what removing AS_PATH length from the BGP route selection algorithm will do to the Internet?
Are you talking about rfc2283 (multiprotocol extentions to bgp)? /vijay
participants (2)
-
I Am Not An Isp -
Vijay Gill