RFC 1955

New Scheme for Internet Routing and Addressing (ENCAPS) for IPNG 🦋

• RFC number: 1955 🎀
• Network Working Group: the group this one lives in 🌸
• Author: Robert Hinden, Ipsilon Networks, Inc. 💅
• Category: Informational (so, not a standard, just an idea with a glow-up) ✨
• Date: June 1996 🌈
• Obsoletes / Updates: nothing listed. It doesn't replace or update any other RFC 🧚

---

Status of This Memo ✨

This memo gives information to the Internet community. 💖 It does not specify an Internet standard of any kind, so it's an idea and not a rule. Distribution of this memo is unlimited, so share it with everyone bestie 🫶

Abstract 🌸

This document was submitted to the IETF IPng area in response to RFC 1550. 💅 Publishing it doesn't mean the IPng area agrees with any of the ideas inside. 🙈 If you have comments, send them to the big-internet@munnari.oz.au mailing list.

This memo describes a proposal that Robert Hinden made to the Routing and Addressing group [ROAD] in January 1992. 🦋 He first sent it as an email. It proposes a medium term solution to the Internet's routing and addressing problems, which is literally the whole point 😭

INTRODUCTION 🎀

Okay so, I want to propose a new scheme that I think is a good medium term solution to the routing and address problems of the internet. ✨ It's giving a lot of pros:

• No Changes to Hosts
• No Changes to Most Routers
• No New Routing Protocols
• No New Internet Protocols
• No Translation of Addresses in Packets
• Reduces the Routing Table Size in All Routers
• Uses the Current Internet Address Structure

Real talk though, it's not a forever solution. 🥺 It does impose some size limits, and it doesn't support new internet services like guaranteed bandwidth, delay, etc. It does make border routers do extra processing, but it doesn't need any packet translation. I think this scheme will buy us enough time to set up a long term solution (like picking one or more of CLNP, *NAT, IDPR, IDRP, Nimrod, Unified, NewIP, etc.) 🔥

This scheme is based on ideas from Deborah Estrin (route on ADs), Martha Steenstrup (encapsulation), and it probably steals from ideas by Noel Chiappa, Van Jacobson, Ross Callon, Dave Oran, and everyone else in the ROAD group. 👑 Iconic lineup, no notes.

CONTEXT 🌈

I think we in the ROAD group all agree that in the short term we need to make way better use of the IP address space. 💜 I also think we (mostly) agree that in the long term we need something that can handle a super large number of end points and routes, and also support new services like guarantees of service and source selected routes. We don't agree on the details yet, but we do agree that we can't figure out a long term solution before March. 💀 We also agree we should start working on long term solution(s) now.

So what's missing is a good medium term solution that keeps the Internet going until we can design and deploy a long term one. 🧚 The medium term solution has to be the most "cost effective" option. It should buy us the most time to build a long term solution, and it should do that with as little change to the existing Internet as possible. ✨

That's why I'm proposing this scheme as the new medium term solution. 💖

NEW SCHEME 🦋

The basic idea is that inter-domain routing gets done by routing on autonomous domains (AD). 🌸 The key is how we do it. The trick is that border routers encapsulate the original IP datagrams inside another IP header. 🎀 The source and destination addresses in that new header (I'll call it the AD-Header from here on) represent the source and destination ADs. 🔥

When the first (entrance) border router gets a datagram from a host or router that doesn't have an AD-Header, it looks at the source and destination addresses and does a DNS lookup to get the addresses for the AD-Header. 😭 Then it adds an AD-Header and forwards the encapsulated datagram to its proper destination AD. Simple, but make it iconic 💅

The border routers would compute AD routes by running a routing protocol between themselves. BGP, or even IS-IS or OSPF for that matter, would work fine. 🌈 As you'll see later, they might even be better. 👑

The addresses I propose for AD addresses are plain old IP addresses. A small number of Class A and Class B addresses would be reserved for this purpose. The network number of the address would show that it's an AD identifier, and the local part of the address would show the actual AD. ✨ This lets us support lots of ADs. For example, 10 Class-A and 10 Class-B addresses could accommodate (10*2^24 + 10*2^16) 168,427,500 ADs. We clearly don't need that many for a long time 🙈

I'd choose to get more than one network number to represent AD addresses because I'd use them to organize the ADs. Let's call these groupings commonwealths. 🧚 Each commonwealth would only have to know the details of its own ADs. Main character organization energy 💜

Next, the border routers would inject these AD addresses into the Intra-AD routing of transit ADs. 🦋 They'd tell the routers inside the transit AD that the border routers are the route to each appropriate AD network. Commonwealths with multiple interconnects (probably the common case) could use subnetting for reasonable routing between commonwealths, as long as they assign the AD addresses carefully. 🌸 This is where OSPF or IS-IS might be better than BGP. Also, IS-IS, with its ability to route on actual end points, might be the best option. 🔥

The reason we inject AD addresses into the Intra-AD routing of transit ADs is that the routers in those ADs can forward AD-Headers without knowing they're special. 🎀 Only the entrance and exit border routers have to do anything different. Teamwork makes the dream work 🫶

Finally, when an AD-Header reaches the last (exit) border router, that router strips off the AD-Header and sends the datagram on to the final destination. 💅

This scheme is built on the idea that IP addresses are globally unique. 🌈 I don't think we'll actually run out of IP addresses for a long time, and we can live with the current addressing until a long term solution is deployed. ✨

This scheme could be extended so it doesn't need globally unique IP addresses. Basically, the combination of AD-Address and IP-Address is the globally unique address. 🧚 To use the scheme without globally unique IP-Addresses and without changing the hosts, you'd need a NAT mechanism in the border routers. 😭 I think it'd be better to change the hosts so they do the DNS query and add the AD-Header themselves. That could be the basis for the long term solution. 💖

Another interesting angle: if we relaxed the current architecture, where one IP-Address always lives in only one AD, and let an IP-Address be in more than one AD, that would solve the issue of letting an IP entity get service from more than one service provider. 👑 Plot twist energy 💜

SUMMARY OF CHANGES REQUIRED 🎀

The DNS needs to be extended to add an AD-Address entry for each name. 🌸 The entry and exit border routers use these AD-Addresses when they build AD-Headers.

Border routers need to be extended to do the DNS lookup, perform AD-Header encapsulation, run an inter-AD routing algorithm using AD-Addresses, and be able to do AD-Header de-encapsulation. 🔥 That's a lot of work, but it's manageable.

CONCLUSION 🦋

I believe this scheme has many advantages. Here's the list, bestie 🫶

• Only border routers and the DNS need to change. No changes are required in hosts or non-border routers. 💅
• No performance impact on datagram forwarding except at entry and exit border routers. 🌈
• Only a small impact on bandwidth utilization on transit networks, due to the addition of a 20 byte IP header to each datagram. ✨
• Removes Inter-AD routing from Intra-AD routing, and as a result solves the routing load (table size and computation) problem for the foreseeable future. 👑
• The routing load on border routers stays manageable, because border routers only need to know the detail of the routing commonwealth they're a member of. Other commonwealths show up as single addresses. 🧚
• No new routing protocols have to be designed or deployed. 💖
• No translation of packets from one address scheme to another. 🦋
• Uses the current IP addressing structure. 🌸
• It scales well even if there's on the order of one AD per IP network, because the AD-Addresses can be assigned logically. 😭

It does have some disadvantages, at least 🥺

• It isn't a long term solution in its initial form.
• It assumes that the current IP-Addresses can stay globally unique for a long time.

REFERENCES 🌈

[ROAD] Gross, P., and P. Almquist, "IESG Deliberations on Routing
       and Addressing", RFC 1380, ANS, Stanford University,
       November 1992.

SECURITY CONSIDERATIONS 💀

Security issues are not discussed in this memo. 🙈 Yes, you read that right. No security stuff here, so read this part with a little caution and don't assume it's been covered 😭

AUTHOR'S ADDRESS 💅

Robert M. Hinden
Ipsilon Networks, Inc.
2191 East Bayshore Road
Suite 100
Palo Alto, CA 94303
USA

EMail: hinden@ipsilon.com
Phone: +1 (415) 846-4604