Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!VAX.FTP.COM!jbvb From: jbvb@VAX.FTP.COM (James B. Van Bokkelen) Newsgroups: comp.protocols.tcp-ip Subject: Re: IP routing and gateways Message-ID: <9007130434.AA06815@vax.ftp.com> Date: 13 Jul 90 04:34:39 GMT Sender: daemon@ucbvax.BERKELEY.EDU Reply-To: jbvb@vax.ftp.com Organization: The Internet Lines: 36 The Host Requirements RFC (lower layers, RFC 1122) has quite a bit to say about this: 1.) Is it the case that I need to provide a default gateway address for IP packets that aren't destined for someone within my subnet? Yes. Allowing several would be better than only one. 2.) Do I need to implement an IGP like RIP and maintain routing tables (even though this device only has one network interface) so that I choose the correct gateway to route my outgoing packets to, or is it OK to go to the same gateway (default) every time? It is quite acceptable to go to the first choice until you fail-over to another in the default list based on upper-layer advice or failure to reply to ARP, etc. ICMP Redirect was designed to supplement the default gateway with real-time advice about which gateway of all those Oactive on the net at any given time is "best". The intent was to allow hosts to avoid hacking routing protocols... 3.) If the answer to 2 is that it's not smart to go to the same gateway every time, then is something like RIP (which should be available on most Unix networks because of routed) the way to go, or does the ICMP redirect take care of letting me know that one internal gateway is better than another for an outgoing packet? The problem with implementing one routing protocol is that you'll have to do them all (RIP, IGRP, OSPF, IS-IS....) before you have real interoperability. In particular, there are many people awaiting their chance to be RIP's pallbearers (at least at their local sites). James B. VanBokkelen 26 Princess St., Wakefield, MA 01880 FTP Software Inc. voice: (617) 246-0900 fax: (617) 246-0901