Just when you thought it was safe to go in the water, things don’t sound so good. We have been hearing about the coming of IPv6 for more than a decade. But the arguments for it have always been a bit vague. All the reasons offered seemed very grand. They have all been something about preserving the “spirit of the Internet” – whatever that is.
But not many of the reasons have been focused on the bottom line. In July 2007, Randy Bush of Internet Initiative Japan and a proponent of IPv6, described the direction of IPv6 by posing this multiple-choice question:
Now they tell us we have to do it because the government says so. Wasn’t that the argument for OSI? We are supposed to give everything a globally visible address to preserve “the spirit of the Internet”? What is this faith-based engineering?
In his comments, Randy Bush then went on to list several things – myths really – about IPv6. Here is what he suggested, followed by my comments in parentheses:
Myth #1: “IPv4 is running out.” (No, it will be around a long time.)
Myth #2: “IPv6 eliminates NATs.” (They wish. Nothing wrong with NATs that a complete architecture wouldn’t fix.)
Myth #3: “IPv6 reduces routing load.” (Quite the contrary.)
Myth #4: “Transition eases routing.” (What transition?)
Myth #5: “IPv6 space is infinite.” (Half the bits are already gone – only 64 left.)
Myth #6: “IPv6 has better security.” (Has the same as IPv4.)
Myth #7: “IPv6 increases battery life.” (This tells us more about marketing than reality.)
Myth #8: “Routers fully support IPv6.” (But not fast. Only in software; no hardware help.)
Myth #9: “No static numbering.” (Transition?)
Myth #10: “IPv6 is deployed.” (In your dreams.)
Myth #11: “IPv6 will replace IPv4.” (At this rate, not in my grandson’s lifetime – and I’m still waiting for grandchildren.)
And then there are at least five problems that Randy Bush doesn’t even bring up or only alludes to. First of all, Microsoft’s transition scheme has already run into scaling problems. Second, there are potential route calculation instabilities inherent in IPv6. Third, there is an inevitability of router table growth due to no multi-homing solution being available (only more patches). Fourth, Moore’s Law won’t bail us out this time (i.e., this is a problem we have known about for 35 years and the solution was published 25 years ago). And fifth, there will be a Rube Goldberg-like mobility scheme caused by using half an architecture. Gosh, I can hardly wait! Paradise truly awaits! Not.
A More Simple Infrastructure
At this stage, everyone should be focused on what they can do with the Internet, rather than patching a 30-year-old design in order to keep the plumbing working. It’s as if we are trying to run the world’s financial markets, hospitals, air traffic control and a myriad of other things while hacking on the equivalent of DOS to keep it running (rather than simply using a complete OS like VMS or UNIX). Oh wait! That’s right! That is what we are doing!
Any new efforts should be towards moving to a complete network architecture; making the infrastructure more simple – not more complex. We should be reducing the “parts count” – not increasing it. We should be solving problems, not papering over them.
“Why is that not happening? Is the Internet architecture running out of steam?” you may ask.
Well, yes. And if it wasn’t for Moore’s Law, it would have happened a long time ago – which actually might have been a godsend. We might have fixed all of this before so many people depended on it.
“But didn’t they see this coming?” you may ask.
Not most of them. Too many neat toys in the sandbox. Their track record at “seeing them coming” isn’t too great.
“Aah, but there is work on a new architecture! The NSF (National Science Foundation) has this big program, right?” you may say.
Well, there is a big program. But new? That is another question. The NSF is Round 2. Several years ago, DARPA (The Defense Advanced Research Projects Agency) funded the big names in the field on a research project to come up with a NewArch. They came up dry.
So now the NSF is trying. But don’t hold your breath. After reading some of the proposals, it looks to me like they are all drinking the same Kool-Aid – and it looks an awful lot like the old Kool-Aid.
“Sounds like a dead-end. So what are we to do?” you may wonder.
Back to Basics
Well, not everyone suffers from groupthink. For a glimpse of what the new architecture can and should be, read my book “Patterns in Network Architecture.” I’ve been thinking about this problem for 40 years. I was working on it in the ARPANET, Internet and OSI (where I was in charge of the infamous seven layer model).
“Aah, so he is bringing OSI back, eh?” you may be asking yourself.
Hardly. Instead I go back to the very basics to see what the problem is really doing. For example, I found that the much-touted dumb network is a delusion. What they have been calling minimal was in fact maximal.
I found, for example, that OSI’s upper 3 layers never existed and that the Internet has only half an architecture. In reality, networking is IPC (Inter-Process Communication) and only IPC – of a single layer consisting of two protocols (that exist and are well-understood but ignored by the Internet) that repeats. This isn’t just some engineering bright idea. Rather, this comes from looking at principles first. Ultimately, this is what we need to be moving to.
The result is a simpler implementation and an architecture that scales indefinitely (in which multi-homing and mobility are inherent in the structure, rather than expensive workarounds that don’t scale). It reduces equipment and operating costs by orders of magnitude. It is inherently more secure (i.e., many common Internet attacks are simply not possible and others are controllable). In other words, a complexity collapse.
It also seems to create a much more competitive marketplace that spurs innovation in telecom, rather than leading to the current stagnation. It opens new possibilities for providers, vendors and enterprises.
“Wow, so it would.”
Right! So you can consider another patch like IPv6 and improve your “Internet spirit.” Or you can consider a solution like “Patterns in Network Architecture” and make improvements to your bottom line. Hmm.
“But we can’t change the Internet! It is too big!” you may say.
First of all, that argument holds up only if you believe the Internet is near the end of its growth. Frankly, I think we have only just begun. It would have been nice if the Internet had had the vision of the early developers. It would have been nice had these problems been solved 30 years ago when we first uncovered them. But we just don’t have that luxury.
Second, the Internet is smaller now than the phone network was back when the Internet first started – and they didn’t let that bother them.
Third, why change it? Let the old Internet be the old Internet. Do new things the new way. Only change old stuff if there is a good business reason to change. There is no need for transition – only adoption. To quote John Lennon, just “Let It Be.”
Since 1984, Day has pioneered the development of network management architectures, as well as several related products and protocols at every layer. A recognized historian, he has also published on the history of China. Most recently, he has contributed to the 2007 Smithsonian Institution exhibit, Encompassing the Globe.
Day is the author of “Patterns in Network Architecture: A Return to Fundamentals.” He can be reached at day@pnabook.com.