kooky BGP tricks

E.B. Dreger eddy+public+spam at noc.everquick.net
Wed Oct 8 08:21:30 UTC 2003


DG> Date: Wed, 08 Oct 2003 09:31:45 +0200
DG> From: Daniel Golding


DG> 1) In a way, its fraudulent

If not mutually agreed in advance between 65000 and 65100, yes.
This is analogous to announcing more specifics from another
provider's space.


DG> 2) Some folks do BGP traffic engineering tricks using AS path. Other folks
DG> do measurement stuff using AS-Path. Therefore, wrong AS paths are bad.

True.  Seeing "1239 $smallprovider 701" would be a bit bizarre,
pretending for a moment that 1239 accepted "$smallprovider 701".


DG> 3) Hacks are unscalable. This is a hack.

In a sense.  In another sense, it's like 65000 sending 65100
partial routes... except it's actually a subset of 65000 doing
the talking.

I should have been more clear in my post.  I originally was
addressing the issue of $smallprovider peering with a transit AS
when $smallprovider's upstream wouldn't.  Ghetto multihoming was
a poorly-added afterthought, and probably muddled the question.
:-/


DG> Why doesn¹t AS65100 just advertise the routes as "65100 65100 i" or "65100
DG> i"? What is gained by doing "65100 65000 I" other than ensuring that if the
DG> link between $downstream and AS65000 goes down, then $downstream becomes
DG> unreachable from AS65000 due to BGP loop detection?

I've always seen inconsistent origin ASNs as "evil".  They work,
but are generally frowned upon... perhaps it is the lesser of
evils, though.


DG> If I were AS65000, I might be disposed to complain to AS65100's upstream or
DG> to the large community about 65100 stealing my AS. I would suspect some evil

Again, I meant to indicate 65000 and 65100 agreed in advance.


DG> intent. Also, if $downstream cant afford $500 for an AS, what makes anyone
DG> think they can afford to pay either upstream?

Very true.

And ASNs are still plentiful; running out is far enough down the
road not to worry...


Eddy
--
Brotsman & Dreger, Inc. - EverQuick Internet Division
Bandwidth, consulting, e-commerce, hosting, and network building
Phone: +1 785 865 5885 Lawrence and [inter]national
Phone: +1 316 794 8922 Wichita
_________________________________________________________________
          DO NOT send mail to the following addresses :
  blacklist at brics.com -or- alfra at intc.net -or- curbjmp at intc.net
Sending mail to spambait addresses is a great way to get blocked.




More information about the NANOG mailing list