From matthew@klinzhai.echo.com Wed Apr 14 08:37:03 1993 Received: from echo-slip.UCSC.EDU by fido.wps.com (5.67/1.34) id AA00845; Wed, 14 Apr 93 08:36:57 -0700 Received: by klinzhai.echo.com (4.1/SMI-4.1.1) id AA26714; Wed, 14 Apr 93 08:35:11 PDT Date: Wed, 14 Apr 93 08:35:11 PDT From: matthew@klinzhai.echo.com (matthew kaufman) Message-Id: <9304141535.AA26714@klinzhai.echo.com> To: tomj@fido.wps.com Subject: Re: CIX packet routing? Cc: matthew@klinzhai.echo.com, randy@psg.com Status: OR If everything i've found out about packet routing is correct... the basic idea is to connect to CIX, advertise routes to all of your networks to the CIX (who then advertises them to Alternet, PSI, CERFnet, and via the ANS/NSFnet to BARRNet and NEARNet and more and more regionals all the time, ANS also learns those routes itself)... this solves the problem of packets getting TO us. you then learn routes from CIX (and you'll end up learning all the networks connected to Alternet, PSI, CERFnet, and some or all of the other regional networks), and send your outbound packets to the CIX if the CIX is advertising a route to there to you, otherwise you send the packet out on your secondary channel (which needs to be any service provider that has a default route pointing to ANS/NSFnet) I'm working on verifying that this is all true, in particular the part about ANS/NSFnet learning routes to all CIX-attached networks from the CIX. (if that part isn't the case, then the inbound routing problem ISN'T solved... hopefully I am correct) And we're also trying to figure out which networks you'd end up needing that secondary channel to reach. (which entails getting a list of what routes the CIX router is advertising today) -matthew kaufman matthew@echo.com From tomj@fido.wps.com Wed Apr 14 02:18:49 1993 Return-Path: Received: from fido.wps.com by klinzhai.echo.com (4.1/SMI-4.1.1) id AA26158; Wed, 14 Apr 93 02:18:44 PDT Received: by fido.wps.com (5.67/1.34) id AA00757; Wed, 14 Apr 93 02:20:16 -0700 From: tomj@fido.wps.com (Tom Jennings) Message-Id: <9304140920.AA00757@fido.wps.com> Subject: Re: CIX packet routing? To: matthew@klinzhai.echo.com (matthew kaufman) Date: Wed, 14 Apr 1993 02:20:16 -0700 (PDT) In-Reply-To: <9304140821.AA26016@klinzhai.echo.com> from "matthew kaufman" at Apr 14, 93 01:21:07 am X-Mailer: ELM [version 2.4 PL21] Content-Type: text Content-Length: 294 Status: R Tim and I could not recreate what it was you told us, re: the (apparent) solution to the CIX routing problem. Do you think you could summarize it for me? I'll pass it to Randy Bush, or you can -- randy@psg.com. -- Tom Jennings / tomj@fido.wps.com / World Power Systems / San Francisco CA From matthew@klinzhai.echo.com Thu Apr 15 11:50:19 1993 Received: from echo-slip.UCSC.EDU by fido.wps.com (5.67/1.34) id AA01363; Thu, 15 Apr 93 11:50:13 -0700 Received: by klinzhai.echo.com (4.1/SMI-4.1.1) id AA00379; Thu, 15 Apr 93 11:48:59 PDT Date: Thu, 15 Apr 93 11:48:59 PDT From: matthew@klinzhai.echo.com (matthew kaufman) Message-Id: <9304151848.AA00379@klinzhai.echo.com> To: gnu@cygnus.com, tomj@fido.wps.com Subject: reply from ANS regarding routing Status: OR I have received one answer from their technical group so far. The following conclusions result: My understanding of the routing was basically correct Connecting to the CIX _does_ solve the inbound routing problem (packets TO us will arrive from anywhere on the net, because CIX routes are advertised to and learned by ANS, so that anyone with a default route pointing to ANS will have their packets forwarded by ANS to CIX to us) (and ANS does _not_ charge CIX on a per-network- number basis for that routing) There are more networks than I thought to which you cannot send packets via the CIX (all NEARnet R & E networks) (because PSI wasn't doing routing preference correctly, ANS stopped advertising NEARnet R&E networks to the CIX) (this is all in addition to those networks which are not CIX members) ANS still considers the connection they have to CIX to be on a "trial basis"... but they'd have a whole lot of unhappy customers if that connectivity went away I will be following up with a service provider or two to see if their default rules would permit connecting at their standard rates but then shipping outbound packets from other organizations via that connection if all goes well, qarin and i should have some time this weekend to write up some preliminary proposal-like things for both the "getting TLG & others a new connection" project and the "connecting santa cruz" project. -matthew