Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!think!ames!ucbcad!ucbvax!UDEL.EDU!Mills
From: Mills@UDEL.EDU
Newsgroups: comp.protocols.tcp-ip
Subject: Re:  Generalized Subnetting
Message-ID: <8708191010.aa11019@Huey.UDEL.EDU>
Date: Wed, 19-Aug-87 10:10:05 EDT
Article-I.D.: Huey.8708191010.aa11019
Posted: Wed Aug 19 10:10:05 1987
Date-Received: Sat, 22-Aug-87 00:47:44 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 16

Phil,

Well, brute force is not enough when there are 270 vanilla IP entries in
the table, not to mention subnets, tunnels and crooked passageways. A
gateway popping 6 ms per packet (133 packet/s) is probably too slow for
any trunkline use. In my experience the depth of search is seldom more
than three, one for the martian filter, one for the subnet and one for
the net. Tunnels and defaults add another one each, but these are seldom
used. With fuzzballs type-of-service adds another level for each TOS
combination and an alternate-routing feature (fallback) adds another. With
these last the tables get so intricate and delicate that I hesitate to
use them outside the lab. This is where we need some clever algorithmic
translation between the handy descriptions you and I are using and the
actual table structure/cache that drives the routing function.

Dave