- il y a 1 semaine
Connectivity
Catégorie
🤖
TechnologieTranscription
00:00Sous-titrage Société Radio-Canada
00:30When you first plugged it in, when you first had that genesis moment, to what degree did you understand that
00:38this was a really, obviously you couldn't have foreseen all that would come, to what degree did you recognize that
00:44this was a really important invention?
00:46Not at all.
00:48At Xerox Research, we were mostly building our own tools, things that we found useful, and we had a new
00:55laser printer and we needed a way to print on it, so we developed this network to do that without
01:01any notion of what would come later as the Internet.
01:05So when did you start to recognize the power of what you'd done?
01:09Well, it took decades. I would say in the 90s when the World Wide Web took off, that's when we
01:18began to see that it might amount to something.
01:20So let's go to the elements of Ethernet that made it succeed, because there were lots of ways of sending
01:26information, and it just so happened that yours became the standard.
01:31Explain the advantages that it had at that moment over its competitors.
01:35Its competitors generally were aimed at serving what was then the dominant method of using computers, dumb terminals, and the
01:43Ethernet was intended to serve personal computers, so that was the key advantage.
01:49They were all serving dumb terminals, and we were serving personal computers, and as you may know, personal computers won
01:57out in the end.
01:59But there were a couple of elements of Ethernet that made it better than its components. Explain a little bit
02:05of the structure of how Ethernet worked that made it succeed.
02:09So three major features. Number one was it brought packets, data packets, the Internet's packets, all the way to the
02:18desktop.
02:19Instead of having them stop at the host and turned into a dumb terminal character mode, the Ethernet brought Internet
02:27packets all the way to the desktop, so you could write all the applications that you wanted that used the
02:32network directly.
02:33The second was that the Ethernet was 10,000 times faster than what it replaced.
02:42Not 10%, not 100%, not 1,000, 10,000 times faster, so that brought in an era of bandwidth abundance,
02:52and then the third thing is the Ethernet was a standard, so the whole Internet movement was driven by a
03:00standards impetus to make it multi-vendor, and Ethernet was part of that.
03:04So explain the innards of how you got it to be 10,000 times better than its competitor.
03:11Well, I'll have to reveal a cheat there.
03:16Please do.
03:16The networks that it replaced were intended to go very long distances, miles, hundreds of miles, whereas Ethernet was designed
03:24to go one mile.
03:26So by cheating, by serving computers within a mile of each other, then we could go way faster.
03:33So you make a decision for optimizing for a shorter distance, but you also have several extremely interesting ideas embedded.
03:40You chop information into different packets, you send packets and you send them out at random intervals.
03:47Explain some of the inner mathematics that was different, because what's so interesting is that it worked because you had
03:55product-market fit,
03:55but it also worked because you had fantastic engineering and mathematical insight that most people were opposed to or disagreed
04:04with at the time, correct?
04:06Yes. So while we were working on it, the computer industry was getting memory.
04:13So the Intel 1103 dynamic RAM chip came out and provided much cheaper memory.
04:22But prior to that, we had to make do without cheap memory.
04:25So we shared a cable, which led to the sharing mechanism that you referred to.
04:32So we were sharing a cable instead of sharing memory for this communication among the PCs.
04:39And so that was the principle.
04:42Then memory became cheaper, and so in subsequent versions of Ethernet,
04:46it moved from sharing a cable to sharing a hub with memory in it.
04:50Oh, I see.
04:51And so that is one of the moments where you suddenly allow companies start to begin to really implement Ethernet?
04:59Yes. Well, there were hundreds of companies at that point.
05:02Right.
05:02And they were innovating on the Ethernet idea.
05:09And so Ethernet had 20 generations since then.
05:13Yeah.
05:14About around this time, you come up with what is known as Metcalfe's Law.
05:18Probably most people here know Metcalfe's Law.
05:20If you didn't know, this is the man who wrote the law.
05:22What I didn't know is that it was a marketing law.
05:26It wasn't a mathematical insight, it was to sell stuff.
05:29So tell the story of how you came up with Metcalfe's Law.
05:32And explain what Metcalfe's Law is for the few people here who don't know.
05:36And congratulations, by the way, on having a law.
05:38It's lovely.
05:38Well, thank you very much.
05:41Our company, 3Com, was selling these Ethernet cards that you plugged into PCs.
05:46And one of the problems were there weren't any PCs.
05:50So we had to push hard to get them sold to the very few that were out there.
05:54So we made a three-node kit that we sold for $3,000.
05:59Three cards that you'd plug into the three PCs.
06:01A cable to connect them all together.
06:04And then some software on a diskette.
06:06And you could share a printer, which in those days was important.
06:10And you could share an email system.
06:15And you could share access to the Internet.
06:18So we sold these for $3,000.
06:21And after a while, our customers concluded that this network,
06:26this three-node network worked like we said it would.
06:28It just wasn't at all useful.
06:31So I came up with a slide that aimed to explain to our customers
06:36why their networks were not useful.
06:39And the slide said the cost of your network goes linearly
06:42based on the number of our cards that you buy.
06:46But the number of potential connections goes up as the square.
06:50So that there's a crossover point out here.
06:53So that the reason that these small networks were not useful
06:56is they were not big enough.
06:59And what was the remedy to that problem?
07:01It was to buy more of our products.
07:05So that chart that I just sketched in 1995, quite a bit later, 15 years later,
07:11was called Metcalf's Law by George Gilder.
07:14Uh-huh.
07:15And I've been defending it ever since.
07:17And whether it was billed for marketing or not,
07:19it turned out to be true.
07:21So...
07:22Well, true to a degree.
07:24There is data that shows that the value does grow more quickly than linear.
07:29And now we debate whether it's the square or the cube
07:32or some other small factor.
07:34I'm defending the square.
07:36You still stand by the square?
07:37Metcalf's Law says the value of a network goes up as the square
07:41of the number of attached devices or people or whatever.
07:44Well, let's go to...
07:45Because this gets to one of the core ideas that you've spoken about,
07:48written about, and think about a lot, which is...
07:50You've argued...
07:51The quote is connectivity is the most important thing that has happened to humanity.
07:56The general idea is that connectivity and the rate of connectivity over the last 20 or so years
08:01is one of the most profound consequences.
08:04Explain the logic, explain what you mean, and work through your theory for all these folks here.
08:09Well, first there's the bold statement that the most important new fact about the human condition
08:14is that we are now suddenly connected.
08:18And when I say suddenly, I mean between 1969 and today, we've added 5 billion users to the internet.
08:26So that's sudden.
08:27And that means with this sudden onslaught of bandwidth, we've had trouble figuring it out.
08:33Yep.
08:33And how to manage it and how to use it.
08:35And so there have been a number of famous pathologies that have emerged from this overwhelming amount of connectivity.
08:44Let me list some of the pathologies.
08:47The first one was hacking.
08:491973 was the first two high school students in Los Angeles hacked the internet in time for Christmas in 1973.
08:59But then next came pornography, and the government of the U.S. almost shut the internet down
09:04because it was so good at carrying pornography.
09:07And so they passed the Communications Decency Act.
09:11And now, as you know, there's no more pornography on the internet.
09:15And then came advertising, which at first was viewed as a pathology,
09:22that people who built the internet were offended that it would be used to carry commercial activity.
09:27But very quickly we discovered that advertising would generate all the income and revenue to support the internet.
09:34Companies like Google famously were built on advertising.
09:37And more recently we have fake news and censorship, which is a response to fake news.
09:45And those are just the latest pathologies, and I'm pretty confident we'll solve them.
09:50So, is the argument that the speed of connectivity, it happens too quickly, so pathologies arise,
09:59but then why does it follow that we can solve the pathologies?
10:02Well, you haven't asked me about the positives, only I brought up the pathologies.
10:08But there are positives of having connectivity, and my point of view is the benefits of connectivity dominate.
10:15So, is the argument, and I wouldn't disagree at all there, absolutely.
10:19You know, the benefits far outweigh the costs.
10:22But of the pathologies that arise, is the argument that the pathologies arise,
10:26but then our collected wisdom, helped in part by connectivity,
10:29allows us to figure out solutions that help us manage them?
10:33Yes, like spam filters.
10:36We invented spam filters because we were getting too much spam.
10:40And so what is the next pathology to arise?
10:42Well, fake news is the big one now, and censorship is a response to that.
10:48But we're going to solve fake news.
10:50And the real question is, who's going to filter our news?
10:55And so is the government going to filter our news?
10:58That doesn't sound like a good solution.
11:01So what are we going to do?
11:03I have proposed that a solution is that we all build our own filters
11:08with tools that make you able to compose multiple filters
11:14to give you the filter that you want.
11:16So a composable filter.
11:20And I propose that as a solution to the fake news problem.
11:23So you propose that as a solution at what level of the internet?
11:26Like at the Google level, as I search through Google News,
11:29inside my social media feeds?
11:30Well, Google could produce a news stream,
11:34and you would have the option of including it in your composed filter,
11:40or not, or some subset of it.
11:43Maybe you have your own criteria that you could apply to the Facebook feed,
11:47or you could make your own feed, and other people could use your feed too.
11:54Anyway, it's an app.
11:55There will be an app for that.
11:56Well, that will be a good next step.
12:00Let's go to the positive sides of connectivity.
12:02What to you is the, tell me a benefit of connectivity
12:05that will be surprising to the folks in this audience.
12:07Reduction of poverty.
12:10There's a great graph that shows the growth of extreme poverty since 1800.
12:16And it goes like this, and this, and this, and then around 1995,
12:20suddenly, for the first time in history, it turns down.
12:25Extreme poverty.
12:27On the same graph is the number of connections to the internet,
12:30and it starts around 1985, and it starts up around the time
12:36that extreme poverty is headed down.
12:39So now I understand the difference between causality and correlation,
12:43but correlation is a good first step toward causality.
12:47So I believe that better connectivity has raised the economic productivity
12:52of the world and reduce poverty.
12:54And as Ethernet, you know, is a core component of this,
12:58because it's the channel, or it's the technology over which all of this flows.
13:03Were there, as we reach the stage, 1995, 97, 99, as it becomes clear
13:09that this is going to profoundly reshape the world,
13:12did you recognize imperfections or choices you had made early on
13:16where you wish you had built the technology different,
13:19or where you saw things and you thought,
13:22gosh, you know, I invented this in 73 for a very different world
13:26than what we have 26 years later?
13:28Or did you by happenstance just really invent something that worked straight through?
13:33Well, there were bugs. The biggest bug in the early days of the internet,
13:37and it persists today, anonymity has a very high priority.
13:43And so from the very beginning of building protocols,
13:46we built the protocols giving anonymity high priority to protect privacy.
13:52Yeah.
13:53But having too much, too much anonymity brings in hackers.
13:57Yep.
13:57So I think that was a mistake. We could have played that differently.
14:00And fake news and many of the other pathologies that you've mentioned.
14:06Explain exactly what you mean that you gave anonymity a high priority.
14:10Well, every, this hurts me personally,
14:15but every ethernet packet has a destination address and a source address.
14:21So the source address is a great opportunity to check where that packet's coming from.
14:25Uh-huh.
14:26You can look at the source address and say, wait a minute,
14:28this packet shouldn't be coming from that direction.
14:32This source address, uh, is invalid for that direction.
14:37And, uh, for the longest time nobody's been looking at these source addresses,
14:40but simply ignoring them, which is hurtful.
14:44Wait, so that, but that's not a failure of the way it was constructed,
14:48that's a failure of the way we've used it and built on top of it.
14:50Well, you could put it that way or you could say we, uh, we didn't use that.
14:56We chose not to use the source field as aggressively as we should have.
15:01Ah.
15:01Were there any other choices that you made early on that you came to think were the incorrect choices?
15:07Well, there were things that looked like bad choices, but they're consequences of evolving technology.
15:12Like I've already mentioned how the early version of Ethernet was during the time when memory was expensive.
15:18Yeah.
15:19So we shared a cable.
15:20But then memory became cheaper, so it made sense to share a hub full of memory instead of a cable.
15:27So, was that a mistake? No, it was a...
15:31It's not a mistake, yeah.
15:32It was a decision based on the conditions surrounding, and as they evolved, then you had to fix the original
15:39design,
15:39which is what happened.
15:40That's just evolution, right?
15:42Yes.
15:43But it's multi-front evolution.
15:45Most inventors don't know that their invention has to be consistent with innovations around it, not just itself.
15:52Wait, and this actually leads to a question that is,
15:54it gets to somewhat of an open source model, closed source model.
15:57You invent it, you build it, you're a graduate student, when you do it, it's you.
16:00It's Bob Metcalf.
16:01As Ethernet evolves, how much is you, how much is other people?
16:06Well, it's been 50 years.
16:09There are probably hundreds of people who have a claim on having invented something to do with Ethernet.
16:16Right.
16:17And at conferences like this, I meet all of them.
16:19They come up afterwards.
16:20Anybody here invent part of Ethernet?
16:22If so, please raise your hand.
16:24There's usually one or two Ethernet inventors in every audience.
16:27That's awesome.
16:28Yeah.
16:28Wow.
16:29And all I can say is thank you very much for contributing, because Ethernet's working well.
16:35By the way, the latest Ethernet being standardized now runs at 800 gigabits per second,
16:44whereas the original one, the first one, ran at 2.9 megabits per second.
16:50So we've gone from mega to giga in 50 years.
16:54That's truly amazing.
16:55When did you know that Ethernet was going to win out and be the defining standard?
16:59I mean, there were plenty of competitors for quite a while.
17:02Well, in the end, IBM was the competitor, and they had a network called the Token Ring, and it took
17:08us 20 years to kill it.
17:13Because IBM was most of the computer market in the early days of Ethernet, so they got to win more
17:19than their share back then.
17:20Quite an advantage, right?
17:21So they have much more infrastructure and a worse technology, and it's not always the case that the better technology
17:27wins.
17:27It's not always the case.
17:29In this case, it was everybody but IBM.
17:31And so why did you eventually win that battle?
17:35There are probably several reasons why we won the battle, but I think the main one was that Ethernet was
17:40genuinely standard.
17:41We put it through the IEEE standards process.
17:45IBM saw us do that and said, oh, customers like that, maybe we should do it too.
17:50Yeah.
17:50But their heart, in their dark hearts, they didn't see.
17:54They weren't sincerely committed to standardization.
17:57So over time, the benefits of standardization overwhelmed them, and the IBM Token Ring, Ethernet's principal competitor, died.
18:07So the lesson is if you want to completely transform architecture, something much better, something much smarter,
18:14and you just got to work and grind and build out until you're a standard.
18:17So cheaper, faster, and standard were the ingredients to the Internet's success.
18:21I like that.
18:22That's good.
18:23All right, our last couple of minutes, let's talk about Bob Metcalf's career because you are, I mean, you all
18:29know,
18:30he must be at least 50 years old since he invented this 50 years ago.
18:34You have not stayed still.
18:35You have taken your life and you've put it into decade-long chapters, more or less.
18:39Briefly explain the six chapters of your professional life and why it is that you think of a decade as
18:46about the right amount of time to do one thing.
18:49Well, I'm 77, so I may forget some of the phases, but I was an engineer scientist for a while,
18:5713 years.
18:58Yep.
18:59Then I was an entrepreneur, founding 3Com Corporation, and that was 13 years.
19:06Became a venture capitalist, 10 years.
19:11A journalist, 8 years, writing for InfoWorld, and I was a publisher.
19:18I had a job just like your job, only it was InfoWorld.
19:21Excellent publication.
19:22Then I just finished 12 years as professor of innovation at the University of Texas, and now I've lost track,
19:28but I'm now starting my sixth career, which is a computational engineer.
19:33So I've taken a job at MIT, and I'm now coding again, and it's very hard because I'm trying to
19:39climb up this learning curve.
19:41Well, this is not what people normally do at 77, right?
19:44The normal track at 77 is not to go back and learn linear algebra and to code.
19:50So explain the choice and why you've made this choice.
19:53Well, I have tended to find happiness on the steep part of the learning curve.
19:57Yeah.
19:58And so I'm definitely there now because I'm trying to learn how to be a competent computational engineer.
20:04But I chose that because I remember I used to be sort of a computational engineer 30 or 40 years
20:12ago, and I remember liking it.
20:13Explain that.
20:14You find joy on the steep part of the learning curve.
20:17So are you somebody who, when do you know it's time to jump onto a new curve?
20:23Well, I always make the mistake of staying on too long.
20:28Uh-huh.
20:28So for example, I left 3Com after 13 years.
20:31I should have left after 10.
20:33It takes a while for you to realize that you've overstayed your welcome.
20:37So there's the bottom of the learning curve.
20:40You don't know anything, and that's no fun.
20:42Right.
20:42But then you get on this learning curve, and that's fun because you're learning things every day, and that's exhilarating.
20:47But eventually, you know almost everything about the field, and then it's time to bail.
20:54And that's the part that's hard to realize is when it's time to bail.
20:58But hold on, because that's also the point where you're able to use your expertise to maximum effect, right?
21:02When you've really understood and you really know how to make 3Com work, right?
21:05when you really understand how to run InfoWorld.
21:08So when do you know?
21:09You're at the top.
21:10You know everything.
21:11When do you know it's time to bolt?
21:15So for example, if you've started a company, it's important to have a board of directors, recruit a board of
21:21directors who can make that decision for you.
21:24Yeah.
21:25It's very rare that someone will come into your office and say, it's time for you to fire me.
21:31I'm not really good enough at my job.
21:33So if you're starting a company, have a board of directors that can tell you when you're done.
21:39All right.
21:40So we'll take that as a good lesson.
21:42So what are you going to do?
21:43You go up the learning curve of being a computational engineer.
21:46What does that allow you to do?
21:48So my plan is to take on a number of engagements as a computational engineer.
21:53And my first one is geothermal energy.
21:56So I've begun simulating the operation of geothermal wells.
22:00And I'm looking for some sort of breakthrough in the configuration of those wells.
22:06And so that's my first engagement.
22:07And did you choose that because it's an interesting application of your passion for computational engineering?
22:12Or did you choose computational engineering because you view geothermal wells as an essential part of the climate?
22:16First I chose computational engineering.
22:19And then I stumbled onto this geothermal thing, which is extremely promising.
22:25And so that's how, so computational engineer and then geothermal.
22:30So interesting.
22:31Okay.
22:31So you're going to learn these skills.
22:32You're going to go up the learning curve.
22:33And as you go, you're going to find applications that have societal benefits like geothermal wells.
22:39Yes.
22:40So you'll do this.
22:40You'll solve this problem.
22:41And then what?
22:43Well, that'll take about 10 years.
22:45Right.
22:46And then I'll do my seventh career.
22:48Don't ask me what it is.
22:51All right.
22:52Well, what an extraordinary man.
22:54One of the things that I know is that you're supposed to leave these sessions when time expires.
22:58Time is expiring.
22:59What an incredible life.
23:01What incredible work.
23:01We should all be grateful for the thing that this man brought into the world.
23:05Thank you ever so much, Bob Metcalf.
23:06Thank you very much.
23:08Thank you very much.
23:09Thank you.
Commentaires