Master Heading

NPUC Homepage 1997 Workshop 1996 Workshop 1995 Workshop 1994 Workshop 1993 Workshop

Martin Haeberli
Audio Excerpt
Changing in Internet Time
Martin Haeberli
Director of Technology
Netscape Communications Company

Ted Selker: I'll stop talking about standards and start talking about changing an Internet time here. One last thing, all of you will find that in your green folders there is a little chit for lunch and after Martin's talk, we will all meander down the hall and there is a cafeteria that will be on your left down about 50 yards. And for those of you that are interested, there is a nature talk about animals in this area and a walk out into the area, you can maybe get bit by a rattlesnake or jumped on by a wolf and who knows. And the rest of us will have lunch.

Martin Haeberli: Well good afternoon, or good morning, actually this is still morning barely. It is hard, on the one hand, to follow Larry Masinter, and, on the other hand, be between you in either a walk or a lunch or whatever it turns out to be. I was thinking about the indirection comment. And this has come up so frequently for me in the last week that I don't know if there is a new joke one can make up about how many levels of indirection does a computer scientist need to solve a problem? Maybe the answer is one; and the reason it is one is because you go indirect to find the answer to the question. What I want to talk about a little bit is the nature of change in the Internet. One point to reflect on is the Internet is global in scope and becoming more so, not withstanding some very real details about the economics of access on nation's that either are islands or want to become or remain intellectual islands. And there is a lot of debate going on about nationwide proxies not only as a form of performance enhancement but also as a form of thought control going on around the globe. I also want to reflect on the nature of change. I have been thinking about change for a long time, and one of the models I have formed about change in many degrees of freedom is, if you have some fitness function, oh, maybe it is transistors per chip, but you might pick any one you like. Many of them these days have the characteristics that the leading edge moves like exponentially, you know. Coefficients vary and chips, we have Moore's law, that says things double every 18 months to two years, and your favorite area might be different. Maybe in cost for LC displays which has a different slope. But the idea is the same. So the leading edge is nice exponential growth.

Martin Haeberli: The trailing edge, the rate at which technology gets adopted or gets abandoned rather, is very slow. If you think about data storage, my guess is there are still paper tape readers being used as a form of data storage for numerically controlled machine tools in this country, right. And that technology was old 20 years ago. In fact, when I worked on the Arpanet the first, 20 years ago, the loading device for our network switches was not the network as you might expect, but the latest improvement in paper tape technology which was fanfold paper tape. Bob Clements at BBN used to like to say that the only thing worse than fanfold paper tape was nonfanfold paper tape. So, I can also ask the question, where is the center of mass of population? The center of mass of population is growing surprisingly slowly. So you get these kind of, maybe a seismologist would say, this is a recipe for an earthquake, right. Because if this increasing gap does disconnect. And then I would say occasionally you have earthquakes. And, from my point of view, the rate of adoption of the Web technology over the last five years has been such an earthquake on fairly fertile ground. Lots of stuff was precharged. We had the Internet technology, we had the beginnings of a global Internet infrastructure. We had lots of personal computers, and then we had some innovations which, lined up with necessary social changes, caused rapid, albeit not complete, adoption of these technologies.

Martin Haeberli: Another observation is that you have heard it said ontogeny recapitulates phylogeny. I did a little research into this in the last week and I can't tell you why, but I do know that strict evolutionary biologists say this is a little discredited. Nonetheless, the idea that we are recapitulating faithful lessons we have learned in the past and have not had the opportunity to forget, seems to me to ring very true in the field of computing. In the field of network computing, we are kind of rediscovering the problems that we ran into in the last three or four generations of computing where we started out 40 or 50 years ago with the precursors of the Univac One.

Martin Haeberli: Worldwide Internet users, according to pick your market survey, might grow to 200 million by the year 2000. I think this number is actually conservative. But you also see a very profound shift happen from the 40 or so million much more realistic estimate users in 1994 from very e-mail centric use towards very Web centric use. And courtesy of Tony Rukowski, now of General Magic, this happens to be Internet hosts attached on some measure, and my guess is this is permanently attached Internet hosts, is gee, doesn't this look like a nice exponential? Gee, he thought about the log/ln chart again, and projects maybe a 100 million machines attached to the Internet by the end of the century. I think, actually, this graph is again conservative. One piece of evidence we have to support that is we published a number in, I believe, May or June that said that we have over 40 million, actually it was over 38 million but the number is larger now, over 40 million users of client navigator on the outside Internet. How do we know? It is not that we have more than 40 million downloads. No, it is we have more than 40 million unique cookies which we have assigned to people who come to our home page. So that doesn't count people who got Navigator that was preconfigured to go someplace else. That doesn't count people who use Navigator behind a firewall, right. And another nice exponential growth chart for Web servers, actual Netcraft in the U.K. does a survey of outside the firewall Web sites, and their graph is somewhat steeper than this, and their current survey today, which I guess is August, says there are over 350 thousand Web sites visible on the outside Internet. It doesn't seem to show any signs of slowing down yet.

Martin Haeberli: One way of looking at this is that if we look at the worldwide growth of personal computers, we have gone to maybe 260 million now. A bunch of those are business PCs. A lot of them are PCs on a Lan, and yet we have seen IP protocols cache out quite quickly to infect more and more of those PCs that are attached to Lans. Five or 10 years ago there was a standards war of another flavor going on in the Lan community around, well what's your favorite protocol? Is it going to IPX if your Novell, is it going to be SNA, and then in the IBM space is it going to be DecNet, is it going to be AppleTalk. And a lot of those things have not completely died because, as I pointed out earlier, the abandonment rate is slow but are becoming much less relevant. Here is another variant of the, for example, again this is just an example of how fast an infection can spread in this precharged infrastructure. If we looked at NCSA Mosaic, you might find a sort of a smaller peak earlier on, but one of the points here is that there has been a slow and steady growth in desktop applications, classic shrink wrap desktop applications like Word and Excel or Lotus, and there has been quite steep growth in netborn infectiously communicated applications like, say, Netscape. Our CEO, like to call this the Netscape rocket. It would be interesting to see what it's curve looks like in the future. So why is this. Well, as I've pointed out, it is partly because of an available global infrastructure, partly because the growth and the capabilities of content. I gave a talk in Washington, DC a couple of weeks ago, and someone asked me, they said it was really undemocratic of Netscape to come up with new tags. You know, how dare you come up with these new tags. And at the time, in retrospect, I wish I had the wisdom to say, or so your saying that Tim Bursley (?), in 89 or 90 when he went from zero tags to the tags that he invented, was then democratic because he sort of did this innovation. So what I think we started out with in the Internet circa 1990, is we had a lot of text and other stuff moving around with a lot of arcane protocols and a lot of arcane social ritual about how that happened. And how many of you here still know how to use the classic text interface to an ftp site? How any of you here would rather have never learned or rather not learn. I can tell you I used to know how and I try to avoid remembering how, but every once in awhile I am forced into it. So we started out with again a fairly rich precharged collection of text and other content, but it was relatively inaccessible even to those who were skilled. We then had the innovative pragmatic insights of Timbol and his team, CERN in Geneva, who said, you know, guys like Vannevar Bush and Ted Nelson and Doug Engelbart and McLighter and others have been talking for years about, why don't we do hypertext. And there were a couple of attempts to do the perfect thing. Unfortunately, certainly at my great cost, and I suspect many of us in this room have learned the lesson that the perfect is the enemy of the good, and I think that Timbol and his team decided that they weren't going to worry about perfect. They were just going to do something. And so they started out just with the idea of links. What a concept. Been around for a long time. They started out with the idea of links. And that already attracted a small bunch of users and they figured out how to let people get at the server software and made a little bit of text oriented client software...and through synergies in a global Internet centric community over the next two years through this kind of potluck, this potlatch, this community of gifs that we have and are still benefiting from on the Internet, we have innovators also as NCSA who said, oh well, gee maybe we could have some graphics. And I frankly don't know where the contributions came from, but my suspicion is that Larry may know more than I do for things like boldface and other just nice looking text. But we went from stuff to linked stuff to richer text to approachable graphics in media where it was no longer the case that you had to figure out how to log on to an ftp site, to figure out whether you are doing binary or text transfer mode, figure out how to convert at 96 times once it got to your machine to decompress and unwrap it, to something where you can click on something and modules, a very real concern John pointed out, that would play, right. By the way, there is some big opportunities in the next five years about dealing with that real concern. And we are beginning to see, I think both in analysis, that there is a lot of stuff that has been going on in the last five or ten years has been about collaboration or interaction, although it hasn't been explicitly structured in a way that we can easily approach or understand or grasp as such, as well as a lot of stuff which looks like collaboration or interaction building up in the next five to 10 years, everything from live video conferences to good institutional social memories that let you search who said what about this topic, and kind of surf that knowledge space. And it succeeded because it enabled a social infection process. People can try things out, people can set up mail lists, people can talk to each other on the mail lists and say, one of the fun experiences I had at Apple is, we were running a Web server. We were using an extension from somebody in Germany. It was missing one feature. Sent the guy mail on a Wednesday. On Thursday morning I had the fix. I mean, and this is what happens every day. It happened in the Common Lisp communities, it is my understanding, when it was growing. It is happening now, not just with one or two communities, but with hundreds or thousands of communities, and this is just the tip of the iceberg. That has been embodied to some extent in both concrete and rapidly evolving set of standards alphabet soup. I'm not going to dive into these here. Larry talked a little bit about the process by which this is made. My suspicion is it is a little bit like legislation. If you love sausage, don't watch it being made. But we benefit from it, and we need to contribute to both the innovation and variation and some of the selection that goes on, both in what we implement and what we choose to use, how we talk about it with other people.

Martin Haeberli: Another aspect which I think is a real challenge for people comes somewhat to the stability point is presence. In this intermediated space, it can be a real challenge to achieve even seven by twenty-four presence much less, yes indeed, you leave college and the stuff goes away. I actually believe there is a social and a business opportunity for one or more global stable archival services. I know Brewster is working on some variant of that, but I don't know how much he has thought through the whole story. And....naming of the piece. But that is not enough. It is an opportunity and need for global presence. You need to be able to get into this anywhere ...... Some people from Japan complained to me about the lack of performance on their outside network, their access across the Pacific. And it seems to me that again that there is at least two or three things that they can do. One of them is work on running better proxy services on shore. Another one is working on, unfortunately, the much more challenging process of dealing with the regulatory and economic constraints of Japanese telecom industry. This constantly crosses languages. The good news is we have in the http protocol, we have the ability. How many people know this? You can configure your browser to say what your preferred language is. How many people know this? That is the good news. That is, how many Web sites know how to deal with it? Very few today. Even if the Web site knew how to deal with it, how hard is it for you to take your rich set of Spanish or German or Japanese or whatever English content and get it converted to be available to your community of interest who speak other languages. It is very hard today, and my guess is that we will see a couple things happen here. One of them is people will start leveraging these very crude cheap desktop language translation tools. They do quick translations but they do something, both on a static basis, converting sites as well as on a dynamic basis converting stuff on the fly. And there are cultural issues. There is the famous story about how the Chevy Nova didn't sell in South America. It is not enough just to get the language translated. Actually there are probably lots of stuff in user experience and human experience about how you design communication, graphic visual interactions. And there is the issue of how you get scaleable network performance dealt with.

Martin Haeberli: I also want to talk about, sort of further changes which are going to continue spreading the set of branches on this exponential growth, this pattern of parallel exponential growth we have talked about. We are seeing the possibility of us moving in manufacturing from custom, where we were maybe a couple of hundred years ago, to wow we have this great idea, mass production, to what Stan Davis and other have begun calling "mass customization." And to me the first example of that is stuff like this. You know, think about a desktop laser printer. It is a kind of manufacturing plant. You feed it paper and toner. You give it CAD drawings, it happens these days to use mostly postscript, and out comes what you want, pretty quick. We're seeing people moving from bits towards the physical. This is a constrained example of the physical. From constrained, again, this is very constrained, towards universal. Eric Drexler, and I will talk about him just a bit more, and others are saying well, what if we could design things and build them on an atomic scale constrained only by the laws of chemistry and physics? It turns out if you could do that, and, by the way, manufacturing stuff that way right now is pretty hard, like impossible, you could do some very interesting things. Ted mentioned earlier that IBM's lab here, or Ashok mentioned earlier, I forget which, that IBM's lab here is working in somewhat in that general space that is working on atomic force microscopy and other related things. But I say that over the next 20 to 50 years, this is going to have a profound change, not just on our society, but on global culture. And from bulk processes where you sort of start out with a saw, or you start out with a silicon equivalent of a saw, which is you code a wafer of silicon with say aluminum, you code it again with photoresist and then you use light and, boy, is that not atomic. It is a very crude process on atomic scale towards atomic precision. Now this is going to mean that what we think of now as software licensing and distribution problems and rights issues and intellectual property issues, is going to move rapidly into other areas, rapidly, okay, on geologic time, move into other areas over the next 20 or 30 years. I think we are going to see this happen first with software and with media that is obviously deliverable in electronic form, kind of direct bits for bucks electronic commerce. Things like music. And there are clear targets of opportunity. But over time, we are going to see it moving toward physical goods. Actually I will be interested in knowing how these badges were made. There is a company in the Bay Area called Laser Custom Design that is working toward the point where you send them a CAD pattern and within a couple of days you have precision cut plastic in your choice of thicknesses. And it is useful today for building models of devices, but what if you had a CAD system that actually let you build mechanical systems like that. What if the degrees of freedom get changed? I know that people at Stanford, Mark Levorian and crew, are working on sort of what they call 3-D lithography where you scan an object here and you make a replica over there. Those are, again, just baby steps. So I have talked, I think, about most of these. These happen to be patterns of automobile parts which were made first in a CAD system and modeled. Second, the model was sent electronically to the manufacturing facility that build them in like a day. And, finally, essentially these turned out to be molds for metal parts that can be used in cars. So the final sort of end point of this transition is Drexler's vision of an assembler, You give it, as it were, a CAD specification of how atoms are to be put together, and they get put together. So in a world where bits are value, these ideas about how you have a business model that works with it become increasingly important. And I know you are going to hear about that somewhat from Ted Nelson and Doug Harper this afternoon. I also want to acknowledge and recognize the work of the people in Japan who sort of coined the word superdistribution, Brad Cox who has just published a book with that title, who has done a lot of work in object oriented programming and economic models for getting software to be distributed in a community in a way that lets everyone who participates in it's integration and distribution recognize some revenue benefit from it. And there is a guy named Robert John Cost who I have never met, but I think I spoke with him once on the phone, who idea was let's play judo with the copyright laws. One of the problems with the computers is it is so easy to copy bits. The copyright laws were written in the context of it is hard to copy bits. So he argued for let's, in fact, acknowledge that it is easy to copy bits and try to make it so that as the bits get copied, the right things happen so that those who made the innovation to contribution synthesis had the ideas get recognized, which includes money. So, an idea about how superdistributions might work, and this is my own crude creation, is imagine that you had kind of a cryptographically sealed thing that somehow ran on a bunch of platforms. It might be implemented in, say, Java, but maybe Java isn't the right language. It would be an object. It could have an encrypted content and it would also carry with it pieces of code that could perform operations like get some information about me, pay for my content, get the content if it has been paid for. And the systems I've seen, however they are implemented, at least philosophically, have some of these characteristics to them. Some of the systems I have looked at a little bit are a company that used to be called EPR, it is now called Intertrust. There has been some work at IBM and elsewhere on things called cryptolocks, and then Mark Stevic of Xerox Parc has written some essays on ideas in this space, including a paper called Letting Loose the Light.

Martin Haeberli: So another area we are going to see a lot of change and growth in the next decade is Mobile Languages. From our point of view, Java is it practically speaking for the next five to 10 years. It is going to have a profound influence on the Net. On the other hand, having been at Apple where we made single processor decisions, we married ourselves in 68000 architecture, and having watched the costs of that in the long run, I also acknowledge that in the long run there is room for innovation and variation here. And we are going to see ...

[inaudible question from audience, following text is speaker addressing question]: So ActiveX is left out partly, in fact directly, as a result of my definition of mobility. The ActiveX substantiations today, as far as I know, don't yet have any of these characteristics and were not initially envisioned to have them. Now I understand Microsoft has articulated that they intend to wrap ActiveX around Java stuff, so in theory you could generate some mobile that is retardable ...

Martin Haeberli: But actually there is a lot of depth here. You think about mobile languages. One aspect is the stuff can run on any platform. But it actually goes deeper than that, and Java isn't yet there. Some of these other languages are moving in that direction. The idea is a computation can run for a while and then resume someplace else. The General Magic guys had made some stabs at commercializing this but there is lots of other thinking going on in this space. Martin Haeberli: So to come back to my them of "Why has the Internet Begun Changing So Fast and How Much Will It Continue to Change Things," my answer is because we've all had the opportunity to stand on the shoulders of giants. I've mentioned some of them before. Some of them are, fortunately, still with us and still very vocal. Doug Engelbart's still charging away trying to change the world and it is really wonderful to watch. Again, there still remain challenges. There are profound social and cultural changes. I am not an anthropologist. I don't know how you design and support and nurture the necessary social changes to make it easier for people to adapt when it is really constructive for human values. And this is not about brainwashing people, this is about making our systems more adaptable to human needs and understanding what social and cultural processes support that. There are some government views about whether this is a good idea or not. For example, finding one extreme by governments who take the our national proxy server will control what you see. To others, including our own, who have perspectives on what kind of technologies may be made available to U.S. citizens and for export to enable private collaboration and conversation whether it is for business or personal reasons. And there are profound issues around taxation. In a world where we move to, oh say a century hence, where everything we want is expressed first as bits. How do we tax it. We have this problem already today. And I am not a lawyer, but I play one on TV. There is the problem that, for example, if I sell you bits electronically with no physical manifestation today, how much tax I should charge you, whether or not, in fact, I owe you tax is subject to some local interpretation. And the mechanisms to actually make that work well are completely absent. The good news is I am beginning to get to the point where I can actually find out based on zip code how much sales tax in theory I would owe to the right place. The bad news is that I am successful and I have 10,000 transactions say, this month, off of my site on an electronically delivered product, I am a small entrepreneur. I might be having to write checks to 5,000 different jurisdictions and I don't even know who they are, right. And I don't want to get into that. So electronic commerce has lots of barriers as well.

Martin Haeberli: Another set of changes that we are going to see is exemplified by this rather pitiful collection of devices I schlep around with me, right. You've got your pager, your cell phone, your pilot; this one is just a fun toy. It's a GPS receiver which, by the way, my main frustration is that it doesn't know where I am right now because it can't see the satellites, right. And then, for example, I didn't know Ted was going to have these gadgets which are wonderful. I want one. I just got this which is a Windows 95 machine which very soon will have a little wireless ethernet and ricochet support, and so this will be yet another form of ubiquitous computer that supports kind of where John is, which is where lots of times I want to read the Web and I am not necessarily in a place where I can be connected using classic copper or glass. So Mark Wiser talks about how, in fact here is a small manifestation, how the ratios of computers if you count processors or even devices per person is going to change rapidly, at least in this country, over the next couple of decades. I also suggest there is an opportunity for a resynthesis. I mean, do I really need to carry all these things? Probably not. So I've talked about some of these things and there are probably some more, and Larry talked about the naming challenge. In the short term, I think we are going to begin seeing some improvements and innovation in the commentary annotation and backlink space on the Web. I know that the people at Forsythe Institute, SGI and elsewhere are beginning to work on leveraging search engines, and this is public data. You are welcome to go look at the Forsythe Web site which is They are beginning to leverage search engines to have a context for looking at commentary on a page. So it might be possible to have buttons either in Web browsers or on Web pages that say who points to this page. And so if that gets to the Web browsers, and I see Danny's page and I like some things about it, I might put up on my Web page, Here is my comments on Danny's cool stuff. And somebody looking at Danny's page could say who said something about this. And they could find a whole list of commentary. Now that is real just a small baby step in the space of social infrastructure, but if you can integrate that into the knowledge the system has about how you want to interact with it and who is in your community, there are some interesting opportunities.

Martin Haeberli: The component software market has been ballyhooed many times in the last decade by various people. I think many of the precursors are here, but we're still not quite there yet. Part of the issue is economic model and part of it is what platform and part of it is, among other things, today this iteration, the ActiveX versus Java debate, but actually I see it is going to be played out in a couple of phases. Within a couple of years, I suspect we're going to have much more incremental netborne software but the majority of software still won't be there as much as I hope it is until perhaps the end of the decade, and we will still probably see more phase changes on beyond Java but I don't know what they are.

Martin Haeberli: There are some wonderful ideas, and Doug Harper will talk to you about some of them, about market-based open systems where computing processes collaborate with each other to achieve their goals or goals imposed on them by humans or human goals by bidding with each other for resources. The wonderfully crazy guy now at CalTech named Robin Hanson who had an idea called Idea Futures. How many here know about Idea Futures? Well more of you can learn about it. The guys in Canada who have been running Idea Futures bidding market have just moved it to a new site. The idea of Idea Futures is what if I could bet on the outcome of a future event or a scientific experiment. And the Idea Futures market is open right now. The bad news is, it's monopoly money, unlike the Iowa presidential campaign futures market which is real money. The good news is it has a very diverse set of propositions and you can post your own propositions and judge propositions and so forth. There was a proposition last year about whether or not Java would win on the Web. There was a proposition last year, for that matter, about whether or not Netscape stock would double in the first day. And basically people can bet both sides of the issues, and you might actually view the current value of the betting equation, yes or no, as a measure of the likely outcome, although, in fact, it might not be the outcome.

[Inaudible question from audience.]

Martin Haeberli: Your question was, didn't there exist a good place to bet on NetScape stock already? And the answer to that, arguably, was no in the sense that people didn't have the opportunity to participate at the IPO, right. And, on the other hand, this wasn't a good place in the sense that this was funny money. But it was an interesting way of predicting what people thought would happen. And then I have talked already, I think, a little bit about nanotechnology. So one way, at least, I have of coping with thinking about the possibilities implied by these changes is by reading. And, for me, there is a set of three or four books which are fun. If you haven't looked at them. One of them, which is in some sense leveraging an idea of a global collaborative communications space, is the book Earth, which may or may not even be available any more. It is a science fiction book. Second is the book Future Perfect I mentioned earlier. The third is a series of books by Eric Drexler and others on nanotechnology. And the fourth, Diamond Age, is a what happens if nanotechnology becomes real; very rich science fiction story.

Martin Haeberli: So that's my prepared remarks. I actually have a quick demo I want to give you in the kind of what's on beyond text and graphics space in the Web stuff. And I will be glad to make this presentation available to the conference organizers so that you can look at it later.

Martin Haeberli: So now we wait for screen redraw, I hope. I hope. No, the problem is not caching, the problem is, okay, well, we'll just see what happens here. Ohhhhhhh, please don't do this to me.

Ted Selker: Can you do this through Netscape over the Internet.

Martin Haeberli: Oh, well yes. You can actually do this over the Internet. This is not an issue of caching. This is, in fact what I'll do, I will talk about the issue while I solve the problem. The fundamental problem, in my view, is that what we are seeing here is a set of interactions with Windows which may or may not be idiosyncratic to us, but have to do with how are color maps dealt with by other applications? Right. And the answer, unfortunately, is not very well. And as somebody who was at Apple for 14 years, I'd like to think that the Mac handles it somewhat better. And my experience is that the Mac does handle it somewhat better. You can also say, well it doesn't really matter. We at Netscape should just solve it for our application. But, and now I am running into the Windows doesn't want to let me launch an application mode. Let's see. Give me a couple more seconds. Why don't we take some questions while I try to get this demo to fly.

Speaker: You and Larry both talk longingly about preservation. What kind of policies do think are appropriate for preservation? I'm not actually sure that all the term papers that I wrote in college I want to still have available, and certainly there are other things I have put out on the various places that probably shouldn't be seen. So this whole issue of having temporary storage and having some notion of a lifetime and some notion of what this is in addition to where it is and how it is sort of not there. You say a few words about how Netscape, does Netscape view itself as being part of that business?

Speaker: We at this time do not view ourselves as being in the preservation business. And, in some sense, I think preservation has at least two aspects. One of them is stable naming, even if it is some form of proxy stable naming. That is, how many of you here know about a trick called server redirect? Okay, server redirect trick is the following: you go to a server and you ask for something. And the server says, one option it can say to you is here it is. Wow, what a win. Another option is it can say sorry I don't have it, go away. Not very pleasant. There is a third option which is built into the http protocol where the server administrator can say if Danny or anyone asks for this thing, well what they really meant to do what go over there. And so the server can tell the client, forget your last request; go over here. And "over here" can be anywhere on the Web, right. So this is a precursor. So one might imagine that you could have a persistent service. If you're a librarian, and you want to have stable bib sites for things. You could have a persistence service that let you use that as your permanent proxy cache. And anytime you got anything over the Web, this thing would remember it for you. And then that service would be somehow run maybe by OCLC or Brewster Kale or someone to be there forever. Now there is the issue, as you point out, about well you published a paper when you were an undergraduate that you didn't want other people to see. Now you're kind of not so proud of it, and you don't want other people to see it. And that also speaks to some copyright issues. If I publish something, the publisher is defunk, who owns the copyright? Who should have a say? So, from a social policy standpoint, my answer to that would be ideally there would be some technologies and some services that would let you go find out what the archival service has that you wrote; I'm waving my hand here completely. And let you say, this is what I want to have happen with that. I don't want people to see it or I do. Now there are questions like how do you establish your authority to actually have that opinion. But ...

Speaker: And the other question is, who pays for that?

Martin Haeberli: Right, so the other question is who pays for it? And Brewster will tell you, if you get a chance to talk to him, about how cheap it is to save the archived things. One of the trends that I didn't focus on is evolution and payment systems. But there is a lot of evolution and payment systems. I think there are a couple of people who could pay for it. One of them is, if I'm a librarian or I'm a writer and I want to cite, say Vannevar Bush's famous article on the Web that was published 30 years ago, because it's of value to me to have my audience have stable access to that, I might pay for it. The fee might be modest. My institution might pay for it because it helps my reputation. I don't think it is necessarily going to be a free service. But actually I think there is a business there. I might pay for it because I want to have somebody that wants to look at this knowledge base, because I want to have the guarantee that anything I've gone to last week, and I appoint somebody else that will be there next year.

Speaker: It is certainly much cheaper now to save everything than to try and decide what to save. The cost of actually making a decision about whether or not it is worth storing something is much higher than just the disk. The disk to hold the present value of the annuity to buy the disk today and replace it five years from now and five years from then and forever is lower than the cost to print out one copy of it on paper. At least at Kinko's. And so really the issue is much more one of naming and indexing. Which things do you want have names to and make indexable and readily and easily accessible. The preservation, if it could be amortized around, the interesting calculation is if you take everything that you read, you can't save everything that anyone might read. But if you actually saved everything that someone did read, if you take everything that you read, and you figure out how much disk that holds and, you know, it's really people don't read that much. So I think it was about $1.50 to save everything that I read a day, which is a little more than it costs to have your shirts cleaned. And that's a little high, but you think as disk prices go down, we'll reach a threshold where it's really cost effective just to save everything so that if there's a global store of everything that anyone read, then each individual doesn't have to save everything that they read in order to make sure that it will be accessible in the future.

Speaker: Let's see. I'm looking back at my notes on your talk and I take away three messages: that the Internet is going to grow exponentially, that Navigator is the platform of choice, and that market is going to move to greater customization and finer grain transactions. And, now these are messages that I'm pretty familiar with already, and ...

Martin Haeberli: By the way, I actually don't want to emphasize the second one. In this audience, I'm just saying as it happens, and here is a measure of how fast things can change in the marketplace. Keep going.

Speaker: Yes, so things can change fast in the marketplace, and I'm wondering how we can take that conversation, which is a pretty familiar conversation, down to a more concrete level. The live discussion focused on archiving because that was one of the areas where the presentation go relatively concrete. But you emphasize that that is not a market you are in. And I would like to hear, I mean, what is the technology vision, more concretely, that is going to respond to these mega trends. How can we get that discussion down to a more concrete level and a more actionable level.

Speaker: Well, I guess, I don't feel like I have really noticed the actual answer to that except to say that this conversation, among others, are good places to begin identifying things. And we talked about the archiving stuff. The stuff this morning about what you can and can't do with e-mail and different points of view about the right things to do with e-mail, I think, are indicative of possibilities. We have talked about the proxy arms race and the style sheet arms race. We are going to see lots of other arms races like that, which isn't necessarily very actionable but it is an observation. On the other hand, there are lots of things that are barriers to adoption by end-users ranging from the complexity of URLs as we see them today to, in some other context, the lack of good archival service which was already touched on, to the lack of kind of persistence. Even people who are running professional Web sites actually, I think, today have a less than adequate vision to be kind and a lousy vision to be honest about what it takes to really run seven by twenty-four services that don't fail. I mean, without beating on ourselves, I think that we've done pretty well, but this week was a hard week for us given people were trying to download our new client and couldn't, right. Last week was a hard week for Microsoft. Last week was also independently and interestingly a hard week for AOL. And there are other people than ourselves who are, for example, in the Web server business who are not yet even running their outside Web sites on our stuff. So, the impact this had on end-users is what really matters in the conversation Ted asked us to have today. And the impact of that is that if you go to travel off sometimes it's not there. If you go to XYZ company, sometimes it's not there. And that means that user doesn't come back. And that means that user doesn't get their friend to use it for the potential benefits. That probably doesn't answer your question.

Speaker: Can I put an annotates, to use one of your words, to answer Phil's question. If you had to bet over the next two years where the bulk of your learning from your customers is going to occur, is it going to occur from your interactions with Intranet customers or is it going to occur in the so-called mass consumer marketplace.

Martin Haeberli: That is a very concrete question which I am happy to answer. And I think, I don't want to discount Phil. I really would like to talk to you at lunch and try to understand how to have a longer conversation. I think, first, for us Intranets are our current business focus partly because, to paraphrase Willie Sutton, thank you, that's where the money is, right. On the other hand, we have customers who use our client navigator and we have customers who use our servers and the tools that go with it which we and many other partners are building on. And, I believe we are going to learn a lot from both people who are users outside the firewall and people who are users inside the firewall. One of the challenges we have, and yet maybe it is a more computerized opportunity in addition to archival is how do we, or how does Chrysler next year or this year, deal with the mass of information that flows to us. I mean, we have 80 million+ hits a day. We have, I don't know what the number is, potentially, especially if we were aggressively inviting, we might see hundreds of thousands of e-mails a day from people who say I want this feature or people who say this didn't work for me or people who say it would be nice if X or people would say will you help me find my friend, Joe. Now how do we deal with that? And how does Chrysler deal with it in the context of I was driving down the road and my car blew up or I had this problem with this dealer. And so there is a profound and very pragmatic challenge which is, how do you manage that information? How do you cluster it? How do you manage it effectively. And there are some companies and technologies that are beginning to move in that direction. That may or may not be more in the space you are talking about. But I think the answer to your question, Michael, is it will probably be a balance. And one of our challenges and opportunities is one, building those things for ourselves

[Audience interruption, beginning inaudible]: ...what practices you steal from the Intranet crowd which is outside the have to prioritize.

Speaker: You ask the best practices. I would suggest that we are ourselves working to capsulate our best practices which show how to run our Web server site, and I think that's learning that comes from our Internet experience which accrues to our Intranet businesses, right. In terms of best practices on software development, I think that is going to come from the Intranet. In terms of what users want, I think it is going to be about 60% Intranet and 40% Internet. But, it's fuzzy.

Ted Selker: We are really interested in what this looks like.

Martin Haeberli: So great, so I'll work on that

Speaker: My question is about two things you said that seem to contradict each other. One is about the complexity in features. And I actually misheard the question about user earnings as user yearnings because I can't afford to save everything I read because the volume is too high, and that just shifts the cost of searching. And, as you said, the programs get more complex and the more you address that by download speed of the program itself, you forget the fact that the program is so complex that not only can I not find the data, I can no longer even find the button that will help me find the data I want. And you know, what sort of global shift are you trying to do to actually try and address that kind of question rather than just keep trying to pour more features and therefore trying to push the slope of that curve up by adding features and weight.

Martin Haeberli: Well, first, Barksdale and Clark say we at Netscape do not have a monopoly on the smart people in the world. We have less than 1500 employees worldwide. And there are lots of other smart people out there. In fact, again, the shoulders of giants point. We and others in this space have gotten to where we are by leveraging other work and nurturing the possibility of other work. So I was actually very excited to see the WBI stuff as something that says "Oh, this might be a baby step in some of the directions you are talking about." I can give you a product pitch. I don't think this is necessarily about our product pitch. But we are profoundly aware of the challenges people have. I don't say we are completely aware, but we are profoundly aware of the challenges people have in navigating this space, both inside corporate networks as well as outside. And we really encourage people to figure out new ways of playing with it. And I think there is going to be a lot of variation in selection. That is kind of the beauty of the cycle that Larry talked about. So, I'll just sort of fly through this a little bit. Perhaps we can get this set up in a demo space. This just happens to be a VR, the VR meld viewer with kind of a progressive disclosure of the planet Earth. And I suffer from, and this also a user interface issue from my point of view, I suffer from not clearly understanding how to fly this airplane, right. And I would like to think that I am not one of the stupidest people in the world, so that suggests that there are lots of other people out there who may even have a harder time learning how to fly this airplane than I do. And what do we do about that? I mean, there are some nice possibilities here that say oh, yeah, gee, you could fly into a view of what California looks like if you knew how to do it. But, you know, where do you get your pilot's license. In fact, I think I just sort of flew through the teleview. What this is suppose to be a demo of, and we can figure out how to make it work a bit better later, I think is progressive disclosure VRML where you basically move as you get closer to something, you get more detail of the underlying space. And you can imagine that if it is done well, this could actually be your access or your portal to an underlying knowledge base. So what you are seeing here is a little bit of the detail on California. I'll stop with that, so thank you.

Ted Selker: As the stars set on Netscape's demo here, I want you guys to think about as you are going down to the cafeteria, as you go into the cafeteria, if you want to there is a nice patio off to the left as you get your food and off to the right there is a hallway with a room labeled 601 if you have signed up for or if there is room for you in this nature talk, there is where you go.

User System Ergonomics Research (USER)
[ IBM home page | Almaden Research Center | IBM Research | (C) | (TM) ]