From osprey67 at yahoo.com Fri Oct 8 14:40:49 2004 From: osprey67 at yahoo.com (Fred Templin) Date: Fri Oct 8 14:41:40 2004 Subject: [rbridge] Re: IPvLX Message-ID: <20041008214049.3841.qmail@web80505.mail.yahoo.com> Hello, On August 23, 2004 I announced to this distribution a -00 document entitled: "IPvLX: IPv6 Addressing in the IPv4 Internet" Since that time, a -01 revision of the base document has obsoleted the -00 version, and a new document entitled: "IPvLX Errata" that specifies patches to be applied against the base document has been submitted to internet-drafts@ietf.org. The documents are currently available at the following URLs (the latter of the two should soon be appearing in the I-D repository): http://www.ietf.org/internet-drafts/draft-templin-ipvlx-01.txt http://www.geocities.com/osprey67/ipvlx/ipvlx_errata-00.txt Sincerely, Fred L. Templin osprey67@yahoo.com -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.postel.org/pipermail/rbridge/attachments/20041008/129302ac/attachment.html From aidan at nicta.com.au Tue Oct 12 20:56:11 2004 From: aidan at nicta.com.au (Aidan Williams) Date: Tue Oct 12 20:57:00 2004 Subject: [rbridge] Minimal ISIS subset? Message-ID: <416CA75B.6050906@nicta.com.au> Hi All, As currently defined, rbridge builds on top of the ISIS routing protocol. Would it make sense to define a subset of the ISIS protocol for use with rbridge? It would seem that an rbridge behaves only as a Level 1 router. - aidan From petera at nortelnetworks.com Tue Oct 12 22:03:39 2004 From: petera at nortelnetworks.com (Peter Ashwood-Smith) Date: Tue Oct 12 22:04:04 2004 Subject: [rbridge] Minimal ISIS subset? Message-ID: I suppose there is value in that for testing purposes since it puts bounds on what has to be tested, and since most vendors would pick up full IS-IS implementations it would help identify exactly what parts of the full implementation to cut out/disable. Peter -----Original Message----- From: rbridge-bounces@postel.org [mailto:rbridge-bounces@postel.org] Sent: Tuesday, October 12, 2004 11:56 PM To: Developing a hybrid router/bridge. Subject: [rbridge] Minimal ISIS subset? Hi All, As currently defined, rbridge builds on top of the ISIS routing protocol. Would it make sense to define a subset of the ISIS protocol for use with rbridge? It would seem that an rbridge behaves only as a Level 1 router. - aidan _______________________________________________ rbridge mailing list rbridge@postel.org http://www.postel.org/mailman/listinfo/rbridge -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.postel.org/pipermail/rbridge/attachments/20041013/25ad72f3/attachment.html From pekkas at netcore.fi Tue Oct 12 22:33:10 2004 From: pekkas at netcore.fi (Pekka Savola) Date: Tue Oct 12 22:34:01 2004 Subject: [rbridge] Minimal ISIS subset? In-Reply-To: <416CA75B.6050906@nicta.com.au> Message-ID: On Wed, 13 Oct 2004, Aidan Williams wrote: > As currently defined, rbridge builds on top of the ISIS > routing protocol. > > Would it make sense to define a subset of the ISIS protocol > for use with rbridge? It would seem that an rbridge behaves > only as a Level 1 router. As I have been commented previously, but without follow-ups, what's the reason to using IS-IS to begin with? None of the enterprise networks I know of (granted, I don't know all too many of them) use IS-IS. Most use either OSPF or RIPv2, if they use a dynamic routing protocol. What's the benefit of IS-IS? I see nothing except it's inherent feature to do both v4 and v6 in the same protocol without hassles if the topology is congruent.. but in enterprises it often isn't, and you need something like draft-ietf-isis-wg-multi-topology-07.txt. -- Pekka Savola "You each name yourselves king, yet the Netcore Oy kingdom bleeds." Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings From Radia.Perlman at Sun.COM Tue Oct 12 23:02:45 2004 From: Radia.Perlman at Sun.COM (Radia Perlman) Date: Tue Oct 12 23:03:01 2004 Subject: [rbridge] Minimal ISIS subset? In-Reply-To: References: Message-ID: <416CC505.2030805@sun.com> 1) Yes. It would be good to say what parts of IS-IS are unnecessary. However, other than level 2 routing (which Aiden noticed), I'm not sure there's much that's irrelevant to RBridges. What IS-IS basically is, is the flooding algorithm and the Dijkstra calculation. So whether the destinations are MAC addresses, IPv4 addresses, IPv6 addresses, or whatever, is very minor. Maybe there is some simplification around the "area addresses". With RBRidges, there should be a single area address, which would be a constant (and not need to be configured).So one wouldn't need the explanation of how to deal with multiple area addresses when two areas are merged (and especially the hairiness around negotiating how many area addresses this area should have). But anyway, we should go through the spec and see what can be removed when just used for RBridge. 2) As for "why IS-IS"? RIP wouldn't be appropriate. We really need a link state protocol. As for OSPF, it requires assigning IP addresses to all the routers, and the format isn't flexible for adding new information, so IS-IS is just easier to adapt. With IS-IS the RBridges can be genuinely zero configuration. Radia Pekka Savola wrote: >On Wed, 13 Oct 2004, Aidan Williams wrote: > > >>As currently defined, rbridge builds on top of the ISIS >>routing protocol. >> >>Would it make sense to define a subset of the ISIS protocol >>for use with rbridge? It would seem that an rbridge behaves >>only as a Level 1 router. >> >> > >As I have been commented previously, but without follow-ups, what's >the reason to using IS-IS to begin with? > >None of the enterprise networks I know of (granted, I don't know >all too many of them) use IS-IS. Most use either OSPF or RIPv2, if >they use a dynamic routing protocol. > >What's the benefit of IS-IS? > >I see nothing except it's inherent feature to do both v4 and v6 in the >same protocol without hassles if the topology is congruent.. but in >enterprises it often isn't, and you need something like >draft-ietf-isis-wg-multi-topology-07.txt. > > > From aidan at nicta.com.au Tue Oct 12 23:33:22 2004 From: aidan at nicta.com.au (Aidan Williams) Date: Tue Oct 12 23:34:05 2004 Subject: [rbridge] Minimal ISIS subset? In-Reply-To: References: Message-ID: <416CCC32.1010701@nicta.com.au> Pekka Savola wrote: > As I have been commented previously, but without follow-ups, what's > the reason to using IS-IS to begin with? > > None of the enterprise networks I know of (granted, I don't know > all too many of them) use IS-IS. Most use either OSPF or RIPv2, if > they use a dynamic routing protocol. > > What's the benefit of IS-IS? > The major benefit is that ISIS does not need configured IP addresses for the routing protocol to work. This helps a lot if you want plug and play rbridges. The IP addresses in RIP and OSPF get used for more than one purpose: 1) to put in headers when sending packets, and 2) to identify the router in the protocol. The IP address identifying the router in the protocol must be unique otherwise all hell breaks loose. Uniquely allocating IP addresses so that RIP and OSPF can be used requires: - manual configuration, or - a unique address allocation protocol, or - a new, unique address format (probably only makes sense for IPv6). A unique address configuration protocol must avoid a chicken and egg dependency on the routing protocol. Dealing with merging distinct rbridge networks looks messy when a unique addr conf protocol is employed. The same IP address could be correctly allocated in both networks when they are apart, creating a conflict when the networks are merged. OSPFv3 uses IPv6 link local for communication, but it still uses a 32bit number (ie IPv4 address) to identify routers in the protocol. On the other hand, ISIS uses a System ID to identify routers in the protocol. The System ID can be automatically initialised from a MAC address. ISIS does not need IP addresses to communicate with routers on the link. Hence, ISIS looks to be well suited for the task. > I see nothing except it's inherent feature to do both v4 and v6 in the > same protocol without hassles if the topology is congruent.. but in > enterprises it often isn't, and you need something like > draft-ietf-isis-wg-multi-topology-07.txt. > Well.. arguably, we don't even need the IP routing features of ISIS for an L2-only rbridge. - aidan From pekkas at netcore.fi Wed Oct 13 00:21:43 2004 From: pekkas at netcore.fi (Pekka Savola) Date: Wed Oct 13 00:22:01 2004 Subject: [rbridge] Minimal ISIS subset? In-Reply-To: <416CCC32.1010701@nicta.com.au> Message-ID: On Wed, 13 Oct 2004, Aidan Williams wrote: > The major benefit is that ISIS does not need configured > IP addresses for the routing protocol to work. > This helps a lot if you want plug and play rbridges. But IS-IS requires one to configure a CLNS address per router, or that's what it requires from us who use it for routing IP at least. And those CLNS addresses look pretty scary to the uninitiated. But maybe this clarifies: > On the other hand, ISIS uses a System ID to identify > routers in the protocol. The System ID can be > automatically initialised from a MAC address. > ISIS does not need IP addresses to communicate with > routers on the link. i.e., I guess you assume that the IS-IS implementation would do such configuration on its own? -- Pekka Savola "You each name yourselves king, yet the Netcore Oy kingdom bleeds." Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings From aidan at nicta.com.au Wed Oct 13 01:18:11 2004 From: aidan at nicta.com.au (Aidan Williams) Date: Wed Oct 13 01:19:01 2004 Subject: [rbridge] Minimal ISIS subset? In-Reply-To: References: Message-ID: <416CE4C3.5070902@nicta.com.au> Pekka Savola wrote: >>On the other hand, ISIS uses a System ID to identify >>routers in the protocol. The System ID can be >>automatically initialised from a MAC address. >>ISIS does not need IP addresses to communicate with >>routers on the link. > > i.e., I guess you assume that the IS-IS implementation would do such > configuration on its own? > Yes. Being a plug and play router is a difficult problem if you have to build on IP addresses that are configured. See draft-williams-ipvlx-ipbridging-01.txt for some thoughts on where the issues are. Ethernet STP/bridging shows that plug and play networking of multiple links can be done -- with the spanning tree protocol (STP) using pre-configured addressing. The point of rbridge is to do a better job than STP does whilst retaining the plug-and-play-ness. - aidan From Radia.Perlman at Sun.COM Wed Oct 13 09:01:33 2004 From: Radia.Perlman at Sun.COM (Radia Perlman) Date: Wed Oct 13 09:02:05 2004 Subject: [rbridge] Minimal ISIS subset? In-Reply-To: References: Message-ID: <416D515D.4060000@sun.com> Pekka said: >>And those CLNS addresses look pretty scary to the uninitiated. Yeah. I think you're talking about the "Area address", and that stuff is turgid to read, but luckily irrelevant for RBridge. We'd pick a constant and say "put this pile of bits into "area address" and nobody would have to understand what an NSAP or Network Entity Title is, or obtain such a thing. I've always thought that it would be nice to rewrite the IS-IS spec and take out the references to CLNS, since it really isn't relevant if you're not using it for CLNS. It might even be nice to do an RBridge-specific writeup of it, which would wind up looking really simple since we could remove all that CLNS stuff and things that got put in and never used (like negotiating the size of IDs within the area). Radia