They had several big jobs to do. The team had to make packet switching work in the real world. They had to turn an ordinary computer into an IMP. They had to write software to control the IMPs. And they had to work with the four host sites to make sure that the IMPs could communicate with their mainframes. The biggest problem was the hardware. 'I'm worried that we won't get the hardware built in time,' said Frank Heart. 'We've done this kind of thing before,' said Severo Ornstein. 'Yes, but there's so much more to do this time,' said Heart. 'We have to design a computer. We have to get Honeywell to understand the design and build it. Then we need to test it.' The IMPs were the heart of the network. Each IMP would stand between a host computer and the telephone system. It would have to translate messages from the host computer into packets for the network. When it received packets, it would have to know whether to build them into a message for its host or pass them on to another IMP. At any moment, all of the IMPs would have to know how the whole network was performing so they could send packets by the most efficient route. Because the IMPs were so important, Frank Heart wanted to make sure that they would never break down. He also wanted them to be impossible to destroy. He imagined students at the host sites opening the IMPs and taking them apart. He tried very hard to make sure that this could not happen. Frank Heart's worries about students were one of the main reasons that he decided to base the IMPs on Honeywell's D D P 516 computer. Honeywell sold this computer to the army. Frank Heart knew that the company had an interesting way of proving that the machine was strong enough to work in a war. So, how do you prove that a computer will not break? To answer this question, Honeywell invited its customers into a large hall. There a DDP-516 was hanging from the ceiling. 'That's interesting,' the customer might say, 'but what does that tell us?' 'Look more closely,' the Honeywell people said. When the customer approached, he saw that the computer was actually working while it was swinging on a rope above the ground. 'That's very good.' 'Oh, no,' said the Honeywell people. 'Not really. But the next thing you'll see is certainly very, very good.' At that moment a tall, strong man walked into the room carrying a large hammer. He swung the hammer, and with a great crash he hit the computer again . . . and again . . . and again. When the computer had stopped swinging, the Honeywell people invited the customer to inspect it again. 'Check and see if it's working now,' they said. It always was. This was almost enough to calm Frank Heart's fears about students. •
The software for IMPs needed to be at least as good as the hardware. The software had to deliver whole messages to the correct destinations. For this, software had to be written that worked even if the hardware didn't — even if an evil student had managed to break one of Frank Heart's IMPs.This is still the way the Internet works today: the software understands how to avoid broken hardware. If a packet does not reach its destination, the software knows. Then it sends that packet again, by a different route if necessary. In the spring of 1969, both the software and the hardware were working in BBN's own building. 'Now we know the network will work,' said Severo Ornstein. 'Don't forget the messages are only travelling a few metres,' Frank Heart warned. 'That isn't a network. We still have to build a system that works over thousands of kilometres.' 'That's true,' Ornstein agreed. 'But we know that the principle is exactly the same if the wire is a metre long or a hundred kilometres long. The phone company says that the length of the wire doesn't matter. It's going to work!' 'I hope you're right,' said Heart.
Гуру Песен