 |
Innovators
Read Transcript | Back to Menu
If you don't see the movie playing,
Download Quicktime 5
Leonard Kleinrock
Computer scientist Leonard Kleinrock (1934-) played
a key role in the development of the Internet. He
developed the theory behind packet-switching, and
his laboratory at UCLA was the home of the first node
of the ARPANET, the forerunner of the Internet.
Transcript:
The way I came to work on packet switching was
rather interesting. I was a Ph.D. student at M.I.T.
looking for a research project, and I was surrounded
by computers. And it was clear to me that not long
from then they would have to talk to each other. And
there was no adequate technology to let computers
talk to each other. The problem was if you used a
telephone network you would tie up a telephone line
100% of the time to send data 1% of the time. When
we are speaking on a telephone conversation we're
silent about 30% of the time, that's OK. When you're
sending data you're silent more than 99% of the time
and its too expensive, to use a dedicated
communication resource that way. So I realized what
you have to do is find a way to share that expensive
data communication line or telephone communication
line to send data. And the way to do that was not to
set up a connection but to let many data
conversations share this communications line at the
same time. The key idea was what I like to call
resource sharing: we're going to share that line.
That's the fundamental breakthrough for data
networks. Packet switching is an example it's a
manifestation of resource sharing. But there are many
other ways in which you could share: store and
forward is one, polling is another, various forms of
time share are a third. So that packet switching
happened to be the technology that caught on but the
reason packet switching was so valuable was
because many people could share the
communications line at the same time. And the line
would not be there waiting for data which was not yet
there. So if I have something to send I'm going to join
the queue and send it as soon as it gets ahead of the
queue. And that line which serves the queue will take
whatever is available to be sent. |
|
|
