Interview with Greg Papadopoulos, Vice President and Chief Technology Officer, Sun Microsystems

Q: I wanted to pick your brain and explore a few issues around how you see what's happening with the way the corporate world is deeply integrating technology. Companies such as GE, GM and Ford are really moving their whole business model to one that's net based. It seems to me there's a bit of an analogy, particularly in the Valley, with the smokestake industries in the 50s, when suddenly the consumer revolution hit them. And it wasn't a question of them producing what they thought was good, it was a question of them finding out a bit more about what people might think was good or wanted, and adapting their self-perception to be closer to consumers.

As I understand your task, you've got the job of understanding that big picture for Sun and working out where to go. Do you think there's been an element of the technology industries driving themselves on the basis of the technologies they create?

A: Oh, yeah. That's a trap you can actually fall into. There's always an element of that. It comes from two sides and the maturing of it is the technology getting shaped more and more by movements in the business model. They interact in the following way.

There are technology fundamentals that become so inescapable that we give them names as "laws". So those are the things like Moore's Law, which is what drove the Valley, and it still does of course. But this incredible march forward of doubling transistor counts every 18 months, or probably what's more interesting is halving the price of a transistor every 18 months. We're probably down now, if you look at modern DRAM, a transistor on a DRAM costs about a billionth of a cent. Basically free in small numbers, right?

So, these laws are things that have these inescapable qualities that if you bet against them, you will lose. For the companies that didn't really think microprocessors were going to happen, or that there were still going to be the discreet designs or mini-computer companies, they just left. You can't bet against those fundamentals.

Actually, it's not a law of physics, it's not a technology property per se. You would look at it and say what causes this thing to double? It's kind of weird! Is it some decay or some isotope! No. It's not that at all. In fact, it's a techno-economic law. It's about this demand-supply cycle of software, the shrink-wrap software cycle, drove this demand for more processing. So I got the next version of the application which needed more processing because it was going to serve more users or had more features or whatever. And so that caused demand for more processing. And that created the opportunity for more software, which created demand for more processing.

Nathan Myhrvold, the former Microsoft CTO, he really describes this beautifully, he's a physicist by training, he talks about the expanding gas of software. Think of the processors of the box…

What happens then is that this cycle is fundamentally a capital cycle, and it's the rate at which you can invest in and capitalise the semi-conductor and the R&D streams back to that to go create the next generation of equipment and processes that shrink it down etc.

So what really has happened, and is making the huge dislocation we are in now, is Moore's Law underneath, but really now another law which is coming into play with a vengeance, which is the law of bandwidth. I call it Gilder's Law, because I got to give him credit, out of respect to George Gilder. This bandwidth cycle which now has a very similar kind of feel to it. It's now that the supply of bandwidth is in a supply/demand cycle with the demand created by network content.

As that content gets richer, it creates a demand cycle for more bandwidth. A characteristic of these cycles is that at any point you have a lot, but it's never enough. Because you kind of want to be chronically undersupplied. Nobody would ever say, I've got enough computing, the world has enough computing we don't need to improve this any more. Even though we have a lot. And the same thing with bandwidth.

So this bandwidth law is one which companies will go out of business if they bet against it. So it has that and is emerging as that property.

( go to Bandwidth Column)

An interesting piece of this is that Gilder's Law has a doubling time every six or nine months. That's sort of a world wide capacity, fibres that go into the ground. Even if you go down to the basic technology, how many bits down a fibre do we get, that's doubling about every nine months, so it's going a lot faster than Moore's Law. So somehow this bandwidth is growing at a rate which is dwarfing the growth rates of semiconductors. That actually has a very interesting implication over time.

But to come back to your main story here, you have a substrata from a set of technology laws that are really integrating over all of the economics that are happening. So, yeah there's all this stuff happening on the business side that's causing all of this other stuff to be happening, but fundamentally they get combined together in this macroeconomic phenomenon that create a set of technology laws. So we'll view that as technology driven.

However, the shift and maturing of the industry is that the kinds of applications that you do with the technology and how you think about it – how you would think about harnessing those laws – are ones that, from what I see, are increasingly driven by how is business getting re-invented. So what does the network do, and what is the effect it has? It's really interesting me that you brought this back to the smoke stack industries and the understanding that there was a consumer out there. Because I see there are two principle phenomena that people get out of the network. They are related concepts.

One of them is a real sense of increased intimacy with the customer, partner, supplier, employee. Far from the network causing the impersonal experience because you don't get face to face any more, it gives opportunities for much more personal interaction, much more one-to-one. The whole notion of portals that are really synergistic with this idea that I go to a website which is mine. If we can pull on that thread, how is my persona, me, going to be represented in the network and how is that going to evolve? What are some of the forces around that and how are companies going to come serve that? Really serve me.

So that's one big thing, and I think companies that don't understand that now the intimacy and relationship of a company to the customer, if they don't get it, they are going to look old and stodgy like the old smokestack industries did that did not see the consumer at all.

It's far more than just putting a catalogue up on a web page. If you take our business as a manufacturer, sure, people want commerce, maybe we'll order our equipment over the net. But then they want to know, well what's the lead time on that. And what's my discount schedule on that. This is not just going to any website, it's got to incorporate the relationship that we've negotiated. What do I get? And then, gee I really do want to check on the status of all the things that I have and I want to look at it different ways. And you know, when I own equipment that I bought from you, Sun ought to know about that equipment. Are there any security holes in it? Because you know when I bought it, you know what's changed and you should know whether I've changed it.

(See related article Bandwidth column)

So there's an intimacy there, that gets in, and that, just like now, if you go to a website that's a storefront and they don't let you order something, you think, boy these folks are really primitive. They're gonna lose. We'll see that continue to escalate, I believe, in the dimension of intimacy, and if you don't as a company continue to figure out how you get closer to what the customer really wants and needs in a relationship, not forcing stuff at them and all this kind of gimmick thing, those people will seem old and out of touch.

Q That has been a really profound step up in expectation just in the last 18 months, hasn't it. Clients becoming much, much more demanding about the type of interactivity they expect.

A It's going so fast. It's going so fast too, for me. You talk about these things, and you think, this will be way in the future and, Gosh! It's now.

The other thing that is incredibly enduring and to me – It's almost a bigger idea than the intimacy, the one-on-one – is community. I think what does make the network different is the ability to build communities like we haven't built before.

I go back to the telephone in this. Fundamentally, the telephone didn't change communications in the sense of we still talk to each other. But what it did, it enabled the one to one interaction in a way that physically would have been impossible before. So, if I wanted to talk to you in Australia, before we had telephones, it was a huge investment of our time, right! To have that conversation we would probably expend about a year of our life just to go and do that, versus now we could expend 15 minutes of our life. The scale was just so dramatically different in terms of the personal investment that relationships could happen over phones that couldn't really physically happen.

So, by analogy, in the Internet, you can create communities that you just couldn't create before. And I see things like eBay as really a community that's created. One that physically couldn't exist. I mean you couldn't get all those people. You can judge the merits of whether it's good, but you just couldn't get all those people together.

Q It existed in a fragmented way in the medium of the newspaper, the classified ads sections of newspapers.

A But the scale, is like, you couldn't print the newspaper thick enough, and people wouldn't have the time to read it, and it has the pictures in there too. So, that technology that created the ability to do the auction and then created the business model. But if you start to follow that cycle. There are small, you know the people who are like artisans who create crafts and things, we always think as society goes on that just sort of gets railroaded and those people disappear. But in fact a lot of them shift over to eBay and they sell their crafts through there and that's their doorway. And it's like, wow, this is unbelievably better than having to go around to all the shows and fairs and setting up their tents.

Or I can go to Aboriginal arts sites in Australia and get access to things that I would have had to have traveled there for before today.

So that actually has this reinforcing thing, that it can perpetuate smaller scale business models that really were not efficient before. Which is sort of an unexpected consequence.

But as I think as this stuff progresses we're now really in a lot of the start-ups that I see, yeah we still see the technology start-ups of here's a new gee whiz database, but the really dominant trend are people who are wanting to project their software as a service. It's a service that the network provides. And as that happens, it's really now a service that based on serving some new business need, or some new consumer need. It's really all of the supply and distribution chains are blown up and now the new intermediaries are coming in and filling in those relationships.

It's a wholesale remaking of business.

Other than that it's nothing to write about!

I don't think you can overhype this.

Q No. But it depends on what's getting hyped though. We had a lot of vapor hype until about April 99. Things that I think are still underhyped are announcements such as those by Ford and GM are their new markets. Saying, let's forget about these vertical supply chain models that we relied upon and fine-tuned over decades to the point where they were pretty damn slick. Now we can create a community model that we'll stick on line. We see that as the opportunity that the technology can deliver to us and we can tell to our suppliers who may not be switched on this yet, hey guys, this is where the world is now.

But the thing that really interests me is when you get guys like Jack Welch and Jac Nasser who are so driven and so impatient, seeing an opportunity like that. They tell their guys, this is what I want, and I'm going to come in tomorrow morning it better be done. And I assume that those guys pick up the phone to you and say this is what we need, and I've got to have it at nine o'clock.

A Yes, right.

Q So will you be able to continue to anticipate the sort of technological demands that …

Yes. I can give you this from just my own personal perspective and my five years at Sun in this capacity at various points in the company, kind of chief scientist, chief technology role, the customer interactions and demands have palpably changed from being ones of explain to us where the technology is going Greg, and how is this evolving and all that, to help us understand how we move our business and what technology do I need to dotcom myself and move forward, and what are the issues of doing it. We have found our company geared toward that, that transition.

It comes back to, you start then inferring lots of qualities about what we have to do for our products and how we behave, which is things like scale, availability, liability, security, integrity of the systems. It sort of massive scale, extreme trust. It's not just that they using the technology to move their business model in some way. If the technology fails, their business stops.

Take eBay for example. It may be a metaphysical question, but where is eBay? Where does it exist? You don't point to a company somewhere. What you do is go down and we'll point to some really large scale computing infrastructure and we'll say, that's eBay. It's right in there.

So the sorts of pressures and duties are that the integrity of that system, under incredibly increasing complexities, you look at the millions of items that they have on auction and the hundreds of thousands of items that they have interacting simultaneously with it, that has to be really really good, and you can't get it really good by being conservative.

Because the other way you fail in this business usually is the success disaster, where you get the exponential of demand, which is no longer limited by bricks and mortar, and so you get this kind of reactive scaling that happens. The word of mouth that happens, the community that gets built so quickly here, you have to be able to ramp up at rates like, like you see E*Trade or Schwab at rates of 10 percent per month.

Nobody ever planned for that in IT before. It's totally outside of the methodical capacity planning process. So it becomes this really intense challenge, which is how do you stay really rapidly innovative. And maybe where the innovation is increasing, because we're going to now have competitive differences by how we innovate on that boundary. And then at the same time be able to offer very high scale and very high integrity in those systems. And that's really where our R&D effort is focussed in the company.

Q SO there's another law in there somewhere that the multiple of Moore's Law and Gilder's Law and the rate of growth of connectivity, that's a kind of activity curve.

A Yes. If you just even take those two laws, their exponential. The issue is that there's an intersection point. And it says that today there's enough computing stuff to keep up with the bandwidth. So we are able to handle the transactions, and the eBay users, and all those things, because all the bits that come over the network that are delivered as a service are being served from something. There are those components, the routing side and the bandwidth provisioning side, but then there's server side. The thing is that these had better stay in balance. The only way I can stay on this curve (the Glider's Law growth) is that I have enough computing to keep up with it.

But it says if I stay on Moore's Law, I just keep increasing the megahertz, going 500, then 800 and then Gigahertz and all that, but I didn't change the scale of the infrastructure, then I would fail. Because I would not be able to keep up with the demand.

The only way that you can snap that (processing power) curve on up, is through scale. I have to put proportionally more and more computing stuff into the network. And scale is an enormous challenge because you increase complexity as you add more things. Before, to serve my site I had this much stuff and now I've got four times that much stuff.

And you've got reliability as an issue. I mean I can tell you something that scales perfectly, which is unreliability. If you stack two computers on top of each other, it's twice as likely one will fail.

So you have fundamentally deal with this. I feel it's so deep it's kind of like I remember reading as a kid, we went to automatic switching in the telephone network because, they said in a very sexist way, that if you had to have the operators plug the plug boards in, you'd employ every woman from age 18 to 35 to do it.

So we had to go to automatic switching, so you could go dial a phone rather than pick up a phone and say connect me to someone. And we're still in, in terms of managing this infrastructure, we're still in pick up the phone and ask someone to do it. It's very person, human intensive to run a site, to project a service out on the Net. Very complex things. And if we continue on that path with our scaling, we'll run out of system administrators or something, we won't have people to do it.

A key aspect of this scale, because if we go out in five years, the divergence may be 1000 to one, I may need a 1000 times more computing and storage associated with the network than I need today, relatively speaking. I have to be able to scale these systems, add more computing resources, without adding to the complexity. That's how we see scale, fundamentally.

So that's pivotal. And then you have to increasingly do it under these levels of, as I say, trust. It has to be available. It's got to be high integrity, you can't lose things, like a bank. And it has to be secure.

Going back to the GE/Jack Welch/propogate down/come talk to us, we get to this phase of, all right, how are we going to look at all of this? What's the new ecosystem for supply chains, everybody has to think of that from their own industry point of view. It seems almost invariably to come back to people coming to a services delivered model that looks like the telephone network in some way.

And then you immediately get into these issues of scale and trust. That has to be fulfilled. Of course there is the expressiveness of the software itself. Increasingly you're getting this phenomenon of in the software industry, we used to sell hardware and an operating system on top. But now people are expecting, well there is a web-server. And in fact there should be an application server, and directory services and messaging. All that's part of the platform too. You can think of it from a platform provider's point as raising the level of abstraction of what we have to promise out to customers. These previously independent applications sediment down into the platform. What we offer is higher and higher levels of capability.

Its actually one of the consequences here, it's the way that you deal with complexity, that people ask us to do more. And we always want to be at the point, from our corporate philosophy, that we don't want to compete with our customers and our partners. So that we're very careful not to be stepping very far into any space that people call an application. But always after, how do applications sort of sediment into the platform that we ratchet up.

That's why we did the iPlanet, Nestscape, Sun alliance. It's very much after this model of what does it take to continue to provide that platform.

Q Are you saying this point of intersection between Moore's and Gilder's laws has been reached now?

A Well, today the network is in balance almost by definition, because it works. The thing is, if you know bandwidth is going to grow like that, Gilder's Law, and you know Moore's Laws' going like that, if you want the network to keep working, you better scale, it's as simple as that.

And I'll tell you there's a fraction of our company that is supplying into this. We moved from this technical workstation company to this enterprise server to this dotcom. And that fraction is hypergrowth. Because it's responding to this need. It's really interesting what it does to our business model.

We have our nice, good hard enterprise business, and we've got this dotcom, networking service provider, ASP (application service provider) thing, just … there's very few times in one's life that you look at these things and say, how big will this get? I dunno, Big! What's the total available market? BIG!

One thing that we won't do. We won't compete with our customers. We are not going to be a service provider. We're an arms merchant! Cause if we went into being a service provider, we'd start competing with our customers, or anywhere close to that being called upon us in order to handle that complexity.

Q This curve gets onto another hobby horse of mine, which is about where the long term research to provide the foundation for this sort of stuff will come from. The Bell Labs, that created the transistor. To quote Gordon Moore, who will feed the monster as those labs are forced to justify their existence more in harder commercial outcomes and companies are under such incredible competitive pressure that they are looking to devote their resources to where they need to keep up.

A It's an excellent question. Yes, transistor out of Bell Labs and laser and all that. Of course the foundation for a lot of that stuff was universities and I think there clearly is a need for society to be investing in basic research. It just can't get handled by industry in any decent way. And a lot of that is just the way the market works, right, in that I take R&D dollars which are expense dollars, and I take basic research which has very, very speculative return. In order for me to be able to justify the basic research inside the company is that I have to have really strong IP ownership of it when I'm done.

And I don't think that's actually a social good. I think there's a much stronger social good to say the society, the US has certainly benefited enormously from the technology industry, which has helped balance the budget deficit and all this type of thing, right. So it makes a lot of sense that society ought to be investing in basic research. That means investing in universities. I feel very, very passionately about that.

That being said, the industrial labs, and their roles and responsibilities, are really ones of being able to take those results and do the kind of synthesis and robustifying and really distillation that really can only happen in the industrial labs. Because there's a certain need to have a relevancy. And we do that. I think the companies that are leading this field have really quite vibrant industrial laboratory component to them.

Even Intel now, for years they've been living off the research of universities, and they've just started their research lab.

I don't see it as somehow industry is just giving up research and we're just living on coattails, but I think society has to come in and recognise that in fact it really has been universities. I mean the Internet. It didn't come from industry. I mean this whole company was founded off Berkeley and Stanford.

Society has to learn that basic research is important and not wring our hands over it, but reinvest in it.

Q As a company, if you think there's an avenue of basic research someone should be pursuing, what do you do about that?

A We certainly stimulate it. And we do a lot of collaborative research, that something we do through our labs sometimes. We don't do the research ourselves necessarily. What we do is, we'll fund it, through universities. And we'll form consortia.

For instance, on the Moore's Law front here, we're really starting to mobilise here in the molecular electronics space. Bringing together some universities, some other companies, and looking at what really is the right investment model.

But that's something that's going to be shared. There'll be companies investing in it, universities and things, but I can tell you molecular electronics is maybe three, five years away. It will help us with Moore's Law, there's no doubt about that, but the basis for that stuff will go back 10, 20, 30 years in academic research. And I can't see that stuff. I only have so many hours in the day. I can't go back and say, you know, this thing's going to be important to Sun in 15 years.

Q But you have to be in a position where you see it when it reaches a point of maturity, you can grab, develop it, get your own IP out of it.

A Right, but I'll tell you one thing that is changing there, that's a progressive aspect of IP generation, that's this whole concept of community development of IP. It's open source, community source are some examples of this. But we really will go in, again, it's an enlightened self-interest, and I think the world is moving. It will be very interesting to watch over the next few years how IP evolves. We really do want as a community to develop it. Everybody wants also to be able to get return for what they contribute, and you'd like to be also able to add value to it. And of course the network, it's going to play a role in that. You see it in the Linux phenomena. Linux only works because of the Internet.

So it's expressing one of the IP forms that are going to result from that. And there will be other ones that have more considered commercial residual rights to them. And we've been working on a model like that which we call community source. Yeah the source is out there, it's sort of available, but there's sort of a residual if people contribute to the IP they can actually share in the royalties from it.

We are going to be working on that IP model, I think the industry will be evolving on those lines.

Q Let me take you back to the relationships with clients. Are you having to find more ways to be talking to them on a regular basis about what are the possibilities that you're seeing coming out of the work you're doing for their business models at the same time as finding out how they're thinking about how they're thinking about their business models or is it part of an on-going relationship.

A No, it's on-going relationships. There are lots of levels of expertise that the company can get over time. People who are out there actually implementing these things, so you grow this depth of understanding in professional services for example, and then you get this dotcom practice that they pull the team in and the start pattern matching. It's like a physician, right! So that's how you build up a kind of institutional memory you are able to execute. At the level of the interplay – is it the technology that drives the model or is it the model that drives the technology – the technology enables some new business models that people go take advantage of and exploit and force it back to new demands on the technology.

Look at an online traders, take E*Trade. The initial architecture of those guys, going from, here is a new technology, which is web-serving, application servers and things, that people could synthesize into a new business, which is online trading. There were a bunch of technologies that had to be in place to do that, but they kind of developed through a lot of people doing technology development.

But now the cycle that develops is, now you're building E*Trade, and you're scaling and you're discovering all your performance issues and all these things, and that drives back to the into the technologist and says I need this scale, I need this performance, I need this reliability, I need these new sets of features. That refines that cycle.

Meanwhile there are people on the technology side that are looking at, okay, I see that and that's kind of a stable cycle, now here are some issues that that cycle is still not solving. I have this new piece of technology. To be concrete about that, it's like, yes we've been evolving all this web server world, and you can do the online trades, but what they didn't solve was the fact that it's really very hard to hook up a machine to the Internet.

So we had a labs group, we did this Genie project, where what we got was radically simplified connectivity.

The typical interaction now (with businesses) is we will get involved with people at really any stage. Somebody comes in and says, I got to do this, I got to dotcom. We'll start an engagement at that level and carry that all the way through to we've got to light up and maintain this.

And that's an increasing demand, that's something we learn as a company, we didn't have to do that before.

But you learn to do it and you get better and doing it.

The things that is interesting to me as we go through that, we're learning, because we see the different business models. There's this kind of positive feedback.

Q Does it create another level of vulnerability for a corporation like Sun? Is the penalty for dropping the ball in one of those corporate relationships higher than it might have been when you were dealing with a technology group in a corporation?

A Absolutely. Because people bet their business on it. It is interesting in both ways. It's an enormous challenge. You really have to be perfect. With literally millions of systems we have out there, it's not going to be perfect everywhere. And we have to deal with that. It's a lot about the process in a company and how you understand. That's really deeply understanding our customers' needs. What really is the expectation on us and how to respond to that. As painful as that is, to take companies through cultural evolutions like this, you start to create, really a barrier to entry. It would be really hard for someone to not only have the great scalable technology but also have the relationship that allows that to come to fruition. But I'm not going to minimise the challenge, because I'll tell you, it's the thing that keeps me awake the most. It's not a Sun particular thing. It's an industry thing. What is it going to take to stay on that scaling curve and to have the really high levels of trust in the systems.

Or are we really fundamentally approaching the system designs. Because as you were talking about people's expectations are growing so rapidly. And where the hype really does hurt is in the short term. It's impossible to over-hype this in the sense that how can you overhype the Industrial Revolution. You know that when you are done with it that everything changed. But it's real easy to overhype what you can do now. So we habitually overestimate the short term under underestimate the long term.

So, you've got to live in that, you have to deal with it, but I'll wake up and think, man, why aren't we building computers like you built airplanes? You go to Boeing, I don't think there are any discussions at Boeing that say, you know, it's okay for our plane to be done for five minutes a year. It's not in the vocabulary. It always works, it always flies. It's a big exception for it not to do that. We will have a lot of scrutiny when one of these things doesn't work.

We are not at that level of maturity. We're close to it in certain things. This is not a Microsoft bash thing, or anything like that, because I've a lot of respect for what that built in terms of a technology model and shrink wrap and all that. But the reliability of that system, it's really embarrassing, I mean collectively embarrassing for us that we really produce this stuff that is fundamentally unreliable. And it's not like somebody could keep working on it for a long time and it would get better. That's what I'm saying. You've got to look at that and you go, there's something broken here. There's something deeply broken here.

You look at and you say, this shouldn't be that way. And that's where I think to me the big leaps, and we're certainly trying to do this, is that it really does become like that telephone in that way. You've got to be able to supply dial tone, you got do that. It's this consumerisation of network access. You can't turn people into computer operators.

Q You see it with SAP. Andersen Consulting is a massive business because of the things that don't work with SAP. But in the last few days suddenly that issue is big news because Westinghouse is saying, sorry shareholders, we bought something that didn't work as we expected when we paid the money. Is that awareness coming as a shock to companies around here? That the awareness that the scrutiny levels are much higher.

A It isn't new news. There have been spectacular failures of people being able to go do the business process re-engineering and wrap the IT systems around it. Many companies made several attempts at that. What you see, for a company like Westinghouse, is that it's an enormous challenge for them. This is a hugely complex company and these are monumental tasks. I don't think you would impugn something bad about Westinghouse. It's really, really hard. We did our re-engineering at Sun and we had a thousand people in the organisation working on this stuff. So those are susceptible to management and execution, and it didn't fit my process, and I couldn't change my process to fit the software, and all kinds of things.

To me that just points to exactly what I'm saying. That the complexity is wrong in some way. And I think, again, why do we come back to talk about the services model. It's okay to have a lot of complexity if I've got my rocket scientists that worry about it and then they can project something out that's a services. So the world telephone network, GSM network, is wickedly complex, but that complexity is not projected out when a company goes and buys telephony. They are not exposed to that complexity. They are not sharing that risk. The risk has been taken by the supplier.

Q Sure, but you still hear people like Guy Kawasaki launching the book that he wrote last year saying don't worry, be crappy. Just ship it and work out the bugs later. I can't think of another industry in the world that would won't be hugely embarrassed even to say that internally.

A I'll tell you the tension here. It comes back to what I was saying about sedimentation. You want to have very rapid innovation and rapid innovation is inherently destabilising. It's disruptive. But as you go through and you discover the value of some innovation, you'd like to be able to draw that in and robustify it and convert it into a good service. And then it requires good architecture of your model. If you look at putting a big wad of software on my desk and ask me to maintain it, the problems is if I start putting innovative pieces into this, because I can't manage the interactions among those things, I get unexplainable interactions among those components. That is inherently a defective architecture. You have to be able to sort of compartmentalise.

Now if I can take you one step back here, if you say, why can I do this better on a network than I can on a desktop. The reason is that the scaling effects allow me to aggregate back. So now I can specialise. So on my desktop, I've got my mail my calendar and my productivity and all of these applications and they all have to share and co-operate with one another and any of them screwing up causing this thing to go weird.

But if I go back to the network and I'm designing a mail system, it does mail. And it serves a million people. And it's complex, but man, we watch it. And here's another one that does office productivity, and we're going to make sure it's right and it works. And we're going to get these different tones, these services, from these independent components. So I've used to scale to isolate. I go back to, it's a Mark Twain quote, the fool says don't put all your eggs in one basket, which is really a way of just spreading around your attention, the wise man says put all your eggs in one basket and just watch that basket!

It's kind of like that right? So, we've got to do email. Let's just let the professionals do email and then project that out as a service, and, of course, that would lead to a whole different type of discussion. Now that I've got that service back end, how does that change the front end and people'' experience. Now I get that, going way back to the beginning of this discussion, now I can actually get the me-to-me intimacy and it's about getting the service when I need where ever I need it at the right time, not being anchored to some piece of technology.