RE: comments on proposed setup

[Rob Gipman <gipmanr@mtn.co.ug>]

Hey but BGP is there to ensure connectivity at all time. This means that bgp
finds out that if the internal link is down the traffic will go external via
the internet. There is no other routing protocol capable to do that for you.

-----Original Message-----
From: Vincent Kaabunga [mailto:kavin@uol.co.ug]
Sent: Monday, April 30, 2001 12:25 PM
To: Rob Gipman
Cc: techies@uixp.co.ug
Subject: RE: comments on proposed setup



On Sun, 29 Apr 2001, Rob Gipman wrote:

> This is not true, the IXP has to have 2 routers and 2 ether switches for
> redundancy and if a route reflector is needed then this can be one of the
> routers or an additional router.

Not necessarily, the IXP in this case is simply  a peering point and the
ISPs maintain their own BGP4 capable routers (which they probably
currently have) and simply peer with anyone of their own choice.

The drawback i foresee is that there is a lot more configuration to be
done at each ISP, but it's inexpensive to setup and requires hardly any
maintenance at the center; and it works. I know, for a fact, that the IXP
in Amsterdam started out this way, as an ether switch in the corner of the
Ripe NCC  offices.

> Ether switches can not run BGP4 unless we
> have an extreme ether switch.

MAy not need BGP4 capable devices in the center to start with.

> Have a look at the IXP white paper Barry Green
> wrote at Cisco. This document was made after investigating IXP all over
the
> world on how they have the setup.

This is IXP as defined by Cisco and the way *they* would want us to do
it.

> p.s. who is Randy?

http://psg.com/~randy

> was he at our meetings?

no.

Later,
	Vincent