[18811] in North American Network Operators' Group
Re: The Cidr Report
daemon@ATHENA.MIT.EDU (Tony Li)
Mon Aug 17 23:38:46 1998
To: alex@nac.net
Cc: nanog@merit.edu, tbates@cisco.com, apops@apnic.net, eof-list@ripe.net,
routing-wg@ripe.net
From: Tony Li <tli@juniper.net>
Date: 17 Aug 1998 20:26:29 -0700
In-Reply-To: alex@nac.net's message of 18 Aug 98 02:26:40 GMT
> > | AS762 118 81 37 31.4% WELLFLEET-AS
>
> Oddly enough, they are announcing the aggregate *and* the specifics:
>
>
> brd1>sho ip bgp reg _762_
> *> 192.32.0.0/16 207.106.96.5 0 4969 6461 1673
> *> 192.32.2.0 207.106.96.5 0 4969 6461 1673
> *> 192.32.3.0 207.106.96.5 0 4969 6461 1673
> *> 192.32.4.0 207.106.96.5 0 4969 6461 1673
> *> 192.32.6.0 207.106.96.5 0 4969 6461 1673
If one were paranoid, one might imagine that they are doing so
intentionally to see how many Cisco routers they can tank through prefix
overload. The problem with this is that a) they aren't also announcing the
/17's, /18's, ... /25's and b) overloading a Cisco seems to mean that
someone buys a bigger Cisco, thereby enhancing Cisco's revenue stream.
One is then left with the theory that they are advertising more specifics
for optimality, but haven't discovered communities yet.
Exercise for the reader: If prefixes of length F and longer are filtered,
and a domain has a prefix of length N, how many prefixes can they propagate
into the backbone? How many before people start proxy aggregating them?
Tony
