Vint Cerf on the Future of the Net

johnd writes "The internet is set to become the basis for just about every form of communication, according to net pioneer Vint Cerf, and he should know what he is talking about. Not terribly in depth, but an interesting read all the same."

Comments

Re:Know what he's talking about?

[cperciva]

vint cerf isn't commercially linked to the internet in the same way that gates is to windows

Err.. Vint Cerf is a senior VP at WorldCom.

One disappointing sentence ...

[jc42]

The Arpanet came before the net and demanded that all computers that connect to it do so with the same hardware and software.

I was really disappointed to read this in the article. First, it wasn't true. There were a lot of such restrictions in the early implementations, but by the time that TCP was spec'd, there were already cases of interconnected hardware and software from different vendors. TCP was a (pretty good) attempt to codify what had been learned about how to do this.

But more important is the point that such single-source restrictions were exactly why ARPA started funding what became the Internet. It was, to a great extent, a response to ongoing problems with electronic gadgets that couldn't talk to each other. The military (and ARPA was a military research agency) wanted this problem solved. What good were all those fancy-schmancy electronic thingies if they couldn't exchange data?

If you look up the early docs from the ARPAnet, you'll see pretty pictures all over the place showing large numbers of electronic gadgets, obviously from a lot of different manufacturers, with lines between them showing the comm links. It's obvious that interconnecting hardware and software from different vendors was a major goal right from the start.

There have been a number of comments on why ARPA gave their development money to universities rather than to commercial vendors. A number of military types were open about this from the start: They had learned that military contractors simply couldn't be forced to work together. Most attempts to get them to cooperate with data comm were pretty much dismal failures. They were competitors, after all. They would pretend to be cooperating, while doing everything they could to fix things so their competitors couldn't cooperate. This is still a problem, of course, and probably always will be. Commercial vendors sabotaging standards is a very familiar process.

So ARPA took the approach of funding an independent gang of academic hackers. Give them equipment and money to pay students to hack away. Fund a few overseers to attempt to coordinate this herd of cats. When they seem to have something working, buy them some fun new hardware and challenge them to incorporate it with the old stuff. Try not to let them get lazy and develop a monoculture of equipment from a single vendor. Watch what they do, and carry off anything they produce that seems useful.

But the intent from the start was to make all electronic gadgets talk sensibly to each other. If the early setups didn't achieve this, it was simply a case of "We're not there yet". The intent was to get there.

Re:Real Time Communication And The Net

[HeghmoH]

The Internet's system of breaking all data up into small, discrete packets and routing each packet to its destination independently is totally different from the traditional telephone system, which creates 'connections' in the system, which take your data from point A to point B. Experience with the Internet is starting to show us that, while packet systems are somewhat harder to get right technically, they are incredibly more versatile and useful. Even the telcos are realizing this as they start to route voice traffic over IP-based systems. So it is entirely conceivable that the Internet will take over for everything, and even though you still have the traditional phone lines, they all just hook into somebody's IP system once they get out of your house.

Re:VoIP

[jc42]

As long as most (older) people I know have a 56k or 64k internet connection, and have to pay per minute online time, VoIP and the like will not become mainstream soon.

Actually, it's happening quite rapidly behind their backs. It's not just Japan that has converted to VoIP. In the past year or so, we've seen the reports here and elsewhere that much of the long-distance and high-capacity lines within the phone system have been silently converted to IP. Here in North America, if you make a call outside your local exchange, there is a rapidly growing probability that it will be packetized and sent over IP (RTP actually) to the other end's exchange, where it will be converted back to analog.

So all those people using 56K modems will have their data converted to analog voice in the modem, converted back to digital at the local TelCo. It will be sent over IP to the remote TelCo, where it will be converted back to analog and fed to a modem, which will convert it back to digital. Each translation will produce a roughly 100-times reduction in bit speed. Yeah, it's tremendously inefficient, compared with just doing IP for the whole thing. That's the way things are done in the modern commercial world.

We still have analog phones in our house. But a couple of years ago, we got a good deal from our cable supplier (RCN) to include phone service over the cable. They installed a little box that connects the incoming cable to the house phones. I asked someone at RCN what this did. The summary was that it "puts the phones on the Internet". I asked if this was what they called "VoIP", and he said "I think so".

It can be difficult to get a straight story in such cases. You may very well be using VoIP at home right now, without knowing it. And the people at the phone company might not know it, either.

Funny thing, in the project that I'm working on now, one thing we're trying to figure out is how to get our text messages converted to voice (solved), and sent out to a phone (not solved). We have digitized voice files, and the computers are on the Internet. You'd think it would be trivial to connect to a phone anywhere there's digital service. But it's far from trivial. Most of the people you talk to within the phone system are interested solely in selling you an expensive "total solution" in which you hand your entire company's data over to them, and can't be persuaded to talk about anything so mundane as delivering a single digital file to a single digital device. Information on how to talk device-to-device is exceedingly difficult to come by.

He is correct (of course he would know too)

[anti-NAT]

If you look at the sentence again, what he is saying is the equipment required to connect to the ARPAnet had to be the same - and it was - all connections were via IMPs or Interface Message Processors. There were different types of computers behind the IMPs

If you use the OSI RM to classify the ARPAnet, as all connections and technology was the same, it was mostly just a big link or data link layer network.

Actually, being more specific, it was sort of like a cross between the network and link layers. It performed routing (network layer characteristic), but all devices were connected together via the same technology (a data link layer characteristic).

The revolution Vint was involved in was the idea of creating a link layer agnostic layer 3 ie. the network layer, hence, that is why he is one of the inventors of IP. Abstracting the network layer functions stopped the network being tied to a link layer technology, which allowed the network to then incorporate links such as satellite links, ethernets, pigeons, two cans and a piece of string, etc. etc. etc. This was not possible on the ARPAnet.

Interestingly, the early model of "TCP/IP" had TCP and IP merged as a single layer. During that development, Vint realised that TCP was a logically separate function to IP, and so it became TCP/IP.

It is also important to realise that the ARPAnet protocols were developed and deployed in the late 1960s, early 70s, where as TCP/IP was developed and started to be deployed in the late 1970s, early 80s. Admittedly, I haven't RFTA quite yet, but it sounds like they may have mixed together the history of the ARPAnet and the Internet, which can cause confusion.

In addition to the few historical documents you can find on the net, a good book on the history of packet switching, the ARPAnet and the Internet is "Where Wizards Stay Up Late".