[15895] in North American Network Operators' Group
Re: consistency and cix
daemon@ATHENA.MIT.EDU (Howard C. Berkowitz)
Fri Mar 27 15:49:11 1998
In-Reply-To: <13595.65005.609119.911733@rip.psg.com>
Date: Fri, 27 Mar 1998 15:18:49 -0500
To: Randy Bush <randy@psg.com>, nanog@merit.edu
From: "Howard C. Berkowitz" <hcb@clark.net>
At 11:41 -0800 3/27/98, Randy Bush wrote:
>V peers with X at a number of points. V and X announce the same routes
>to each other at all these points.
>
>V and X are also members of the CIX and are announcing routes to the CIX
>router. X announces different routes to the CIX router than they
>announce directly to peer V because Y pays X to specifically announce
>Y's routes at the CIX.
Where did Y come from? A Nike commercial? Y ask Y? :-) Personally, I
find numbers easier to follow than letters in these examples, don't U? Is
the multihomed case W?
Slightly more seriously, a little more detail would help.
Is Y's space independent from both X and V? I could see V being unhappy if
X is advertising a more specific route in V's space, although this could be
reasonable if coordinated.
Does Y advertise its routes at any points other than through X? Does Y
have a distinct ASN?
If Y is being advertised as part of X's AS, that seems to be a valid local
peering policy.
>
>Should V be unhappy with X's inconsistent route announcements? Should
>X's announcement via the CIX be consistent with their announcements at
>the other points V and X meet?
>
> o yes, because the CIX is a peering point, though router-based (this
> in itself may be worth a different discussion).
>
> o no, because in realty V and X each are paying customers of the CIX,
> and the CIX is merely announcing their customers' routes to each
> other.
>
Is the underlying issue that someone should announce all routes at an
exchange point? It sounds as if V is complaining because X is selling
transit to Y.
Howard
