Researchers Scheming to Rebuild Internet From Scratch

BobB writes "Stanford University researchers have launched an initiative called the Clean Slate Design for the Internet. The project aims to make the network more secure, have higher throughput, and support better applications, all by essentially rebuilding the Internet from scratch. From the article: 'Among McKeown's cohorts on the effort is electrical engineering Professor Bernd Girod, a pioneer of Internet multimedia delivery. Vendors such as Cisco, Deutsche Telekom and NEC are also involved. The researchers already have projects underway to support their effort: Flow-level models for the future Internet; clean slate approach to wireless spectrum usage; fast dynamic optical light paths for the Internet core; and a clean slate approach to enterprise network security (Ethane).'"

Re:The Six Million Dollar 'Net.

[KingSkippus]

Now if we could only rewrite Windows from the ground up

Didn't you see the story the other day?

We are [reactos.org].

Re:What are the odds

[griebels2]

The problem of IPv6 is due to the fact that it just doesn't work besides IPv4. You essentially need to build and maintain two seperate networks. Yes, you can share the same equipment, but the amount of configuration involved almost never justifies the efforts in corporate environments.

In my opinion, there are a lot of things that need to be fixed for an "Internet for the future". One of the biggest hurdles of course is the address space shortage of IPv4, but there are a lot of other issues which need to be solved. Just to name a few:
- More flexible routing of unique identifiers (let's call them IP numbers), so I can take my "identifier" with me (think mobile phones)
- A solution to the ever growing "global routing table" (BGP4 as it is used today)
- Better support for quality of service from end-to-end.
- Better "multicasting" support, also end-to-end. (Let's avoid burning down networks during "cataclysmic" events)
- Better redundancy. Although dynamic routing protocols should heal this problems, in practice they often fail to do this. Especially in cases where connections are semi-dead)
- A much better built-in protection against DDoSes and other kind of abuses.

Unfortunately, IPv6 really fixes none of those problems, except the IP number shortage. IPv6 also comes at great costs, since you need to upgrade your whole infrastructure at once, or it isn't really usable.

So, IPv6 might have been a nice lesson for the next generation "IP protocol". IMHO this next generation should take the following things in mind:

- Transition only works if it plays nicely with the legacy stuff during the transition.
- Transition has either to be cheap or must have so many advantages that you simply cannot refuse.
- Vendors need to agree upon a single standard, or somebody with a large impact should "dictate" it in the worst scenario.

Reading TFA, I was quite disapointed, because anything about how this transition to this cleanslate network seems to be absent at this time. But it is still a research project and maybe somebody did learn something from the IPv6 "fiasco".

Re:Rebuild the Internet

[Jeffrey Baker]

I'm with you. These guys are completely on crack. Haven't they ever read "Netheads vs. Bellheads"? You do not want to have intelligence inside the network, ever. Intelligence belongs at the edge. The core should be application-unaware, stupid, unreliable, and as simple as possible. Which is the Internet we have today, and it works great, thank you very much.

Re:The Six Million Dollar 'Net.

[GringoCroco]

From the whitepaper PDF:

It should be:
1. Robust and available
2. Inherently secure.
3. Support mobile end-hosts
4. Economically viable and profitable.
5. Evolvable.
6. Predictable
7. Support anonymity where prudent, and accountability where necessary.

Not exactly

[mengel]

I couldn't help chuckling as I read the above post, as it outlines all of the things that were presented as benefits of moving to IPv6 when it was initially released. For example:

• There are several mechanisms for running IPv4 and IPv6 side by side, and that was a major part of the discussion in the IPv6 rollout early on. Medium sized chunks of the net were running IPv6 [6bone.net] for quite a while, and were routed in and out of fairly seamlessly. transition mechanisms were designed [tascomm.fi], long before IPv6 was adopted by the IETF. (the linked RFC is from 1995).
• IPv6 designers also put in tools designed to provide for mobile endpoints, although better designs have come out since.
• IPv6 provides and uses multicast addresses as part of it's initial design, and its multicast is being used [cisco.com] successfully.

You can claim that the implementations provided weren't good enough (although I'd like to see some actual data to back that up), but in fact the folks that did IPv6 did have all of those goals in mind when they put IPv6 together.

Re:The Six Million Dollar 'Net.

[mr_mischief]

It's a reference to an old TV show, "The Six Million Dollar Man". Lee Majors played Steve Austin, who got terribly injured. The Powers That Be rebuilt him as a bionic wonder able to do things no mere flesh-and-blood human can do. The doll^H^H^H^Htelve-inch action figure had a vamera view-finder type of thing that you looked through his head to use (his bionic vision), and a cool karate chop action arm.