Dr. Dobb's Journal September 1999
This month's jeremiad will jump and jag from Java, JavaScript, and Jini to Jobs, Joy, Jacquard, and the Japanese Bill Gates, not to mention Mr. Jones. I will mention Mr. Jones, but briefly, as befits today's peripatetic pace of progress, which places the present PC price point at a paltry pittance and turns June's jim-dandy jimcrack into July's junk, the jetsam of the Jetsons.
That is to say, there doesn't seem to be a whole lot of stopping and smelling the roses going on these days. I'm as guilty as anybody: Most of my writing is in the rose-foregoing 300-to-3000-word range. Drive-by analysis for a fast-lane world. But I occasionally write longer pieces, and the accompanying text box (see "Pirates and History") is all about my reaction to what happens when a long-form work gets repurposed for a short-attention-span medium. This preY2K year, it seems to me, is an occasion for thinking about short-term thinking and its consequences. If you think long enough about the consequences of short-term thinking, you shortly start thinking about long-term thinking, and about its consequences, and about how maybe we need more of it.
In part, this month's column is about the possible consequences of taking a long view of technological progress; that is, longer than next quarter. First, though, I'll digress briefly into a discussion of the novel -- that long-view fictional genre. And I'll digress midway through the column to talk about this year's JavaOne conference, digressing from that to mention a significant Java event that took place outside the conference. If I seem to be taking a long time getting to the point, maybe that's partly the point. (Then again, maybe my attention span is simply so shortened by living in late-20th-century America that I can't wend my way through a few hundred written words without two or three digressions. If so, that's the point.)
I do think we ought to take the long view occasionally, but I say, let's not be in a hurry about it. I call this the long view of the long view.
Literature, Ezra Pound said, is news that stays news. That's if it stays literature: Many novels last only a season, and deserve less. But what about the novel itself? Is the novel, or for that matter the linear narrative, dead or dying, soon to be replaced by some new artform born in the interactive, hyperlinked, multidimensional realm of cyberspace? Academics and critics wonder, "Will the Internet kill the novel?" Personally, I don't think that linear narrative is going to go away any time soon, but how about the novel? It's a frequently forgotten fact that the novel is a relative novelty, a recent literary invention, not over a couple hundred years old. Scroll back to 1749 and Henry Fielding's Tom Jones, and you're pretty much at the dawn of the novel.
So what did Fielding think about interactive fiction?
It turns out that he probably would have embraced it if the publishing technology of his day had just supported it. (It didn't, of course. This was almost a century before Charles Babbage didn't quite build the first computer, and even 55 years before Joseph-Marie Jacquard first incorporated programmability into a mechanical device with his software-controlled loom. Java junkies will appreciate that this was the first recorded use of threaded code.) Fielding would have embraced interactive fiction? I think so. Here's the first sentence of Tom Jones:
An author ought to consider himself, not as a gentleman who gives a private or eleemosynary treat, but rather as one who keeps a public ordinary, at which all persons are welcome for their money.
A "public ordinary" is a public house, or publick house -- a pub, a tavern. "Eleemosynary" means charitable. Fielding, who was one of the creators of the novel, declares in the first sentence of his signature work that an author ought to think of writing as more like managing a noisy bar than producing and delivering a work of art to a grateful, and passive, public. You take a 250-year perspective on the novel and you find that one of its creators would have preferred it to be more interactive. As Alanis Morissette would say, isn't it ironic.
Ironic undermining of one's own central thesis is one of my favorite ways to begin any discussion that could get religious. Like economics. The other implication of Fielding's advice to writers is that he expected to get paid for his efforts. Open-source software, a late-20th-century Fielding would say, should not be an eleemosynary treat. Taking the long view on making money may or may not be a good idea. Take the increasingly common Internet business model of selling at a loss but making it up in volume, for example.
Well, isn't that the strategy? Whatever the strategy is, it seems to be founded on the belief that the laws of economics will be overthrown at some future date. The business plans for Internet startups make interesting reading, and seem to provide all the evidence one could want that the novel is not dead.
I read recently that Masayoshi Son, the Korean-born entrepreneur who has been called the Japanese Bill Gates, is working from a 300-year business plan. A week later, when I went back to the site where I had read the article; it had disappeared. I wonder if the plan itself is any less ephemeral. What medium do you store a 300-year business plan on? Not on mag tape, floppy disk, DIVX, or CD-ROM, certainly.
Most of the documents being produced today are being recorded on media that will not exist in 300 years. Danny Hillis: "When we finally shut down the old PDP-10 at the MIT Artificial Intelligence Lab, there was no place to put files except onto mag tapes that are by now unreadable. So we lost the world's first text editor, the first vision and language programs, and the early correspondence of the founders of artificial intelligence." Stewart Brand: "Science historians can read Galileo's technical correspondence from the 1590s but not Marvin Minsky's from the 1960s." Those historians may look back on our time as a Dark Age, our records all written on the wind, lost to history.
My inspiration for writing about the long view is a book, The Clock of the Long Now: Time and Responsibility by Brand, author of The Media Lab, among other credits. (As I said in July, Brand keeps popping up in the most interesting places. Here he is again.) This, his latest book, was published this year by Basic Books (ISBN 0-465-04512-x). Both aforementioned quotes are from the book.
Brand thinks that the time is ripe to promote the long view and, along with Danny Hillis (designer of massively parallel computer systems and currently working for Disney) and some others, has started a foundation, the Long Now Foundation, to do that. The book presents the case for the long view and describes a 10,000-year project conceived by Hillis. We live in a short-horizon world. Hillis says that when he was a boy, the future was the year 2000; today, the future is still the year 2000. He's seen the future approaching at the rate of a year per year.
The perception may mirror some underlying reality. Graphs of many trends go asymptotic sometime not long after the year 2000. Various variations on Moore's law pile up to form what Brand calls Moore's Wall, or the Singularity. The point beyond which things can't just go on the way they have been -- the point beyond which attempts to predict fail. It's not just that the predictions fail; it actually becomes impossible to extrapolate from past trends to some imaginable future.
That's what I get from reading Brand, anyway. It all sounds very apocalyptic and mystical, but ask yourself: What will the world be like in 50 years? The pace of change is so rapid that predicting the future really is problematic, in a way that it wasn't 50 years ago. Despite the increasing difficulty of prediction, we need more than ever to take a long-term view. Brand points out that different values emerge from different perspectives on time. A commercial perspective undervalues redwood trees and rain forests not because it is evil, but because the timescale natural to commerce is not the time scale natural to nature. We need both perspectives. We need at least six different temporal perspectives, Brand says, from the flash of fashion and commerce through the mid-range perspectives of infrastructure and governance, to the long-range perspectives of culture and nature. Mostly these days, we seem to get by with the first two.
Danny Hillis's clock project is a way to focus attention on what Brand calls the "Long Now," analogous to the way that first view of the earth from space focused attention on the Big Here. Once we could see the earth from space, it was impossible to see Here the same way again, Brand says. (You may recall that he put that picture of the big blue marble on the cover of one of his Whole Earth Catalogs.) Hillis wants to do the same for our view of Now.
His plan is to build a large, perhaps Stonehenge-sized, clock in the desert somewhere, far from the pollution and disruption of cities but still accessible. The clock will be a visible and tangible reminder of the passage of time on different time scales. As you walk through the works of the clock, you see the mechanism that ticks off days, the clock's smallest time unit, on up to the 25,784-year cycle of the precession of the equinoxes. This clock is intended to keep accurate time of that latter time period.
Hillis's design goals for the clock when he started the project were: longevity, maintainability, transparency, evolvability, and scalability. The 10,000-year timeframe creates some interesting challenges in trying to attain these goals. Over the course of 10,000 years, the clock will probably need some maintenance. What technology can one assume will be available in the year 4999? Will clock mechanics in the year 6999 be able to get spare parts? As for transparency, in 10,000 years, somebody is bound to lose the user's manual and the repair manual, so the device had better be its own documentation.
When Hillis tried to figure out how to keep the clock accurate over that long time period, he found no solution that was sufficiently: 1. accurate, 2. reliable, 3. easy to measure, and 4. low-tech (piezoelectric oscillators were judged too hard to maintain over such a time frame). One insight that comes out of this is that maybe the only way to control the future is by forming a partnership with the people who will live there.
I read Brand's book one long afternoon while broken down in Arbuckle, California, on the way home from the JavaOne conference. Three days before, I had listened to Bill Joy saying that Danny Hillis's goals in creating his clock were pretty much the goals behind the development of Java.
This is what is known as The Big Spin, and would more appropriately have been delivered by Scott McNealy.
With all the sessions on technologies promoted by Bill Joy (Java, Jini, JavaSpaces), the conference was a bit of a Joyride, but for a while, JavaOne looked like PalmOneHalf, with huge lines of attendees queued up to buy Palm Vs at a ridiculously low price. This being the second year that cheap Palms were featured could account for the large number of attendees, I suppose, but it's at least as likely that the attendance was proof that Java is mainstream and here to stay, if we had any doubt. The Palm connection underscored the main message of the event -- that Sun was aggressively pushing Java for nonPC devices. What wasn't being emphasized was that this represents a retreat on one front: Connected devices are in, but yesterday's dream of zillions of Java-enabled thin PCs is out. Sun's slogan "The network is the computer" still works today, but in a few years, if the dream of Jini-connected devices plays out, will some of Sun's customers respond, "What's a computer?" I heard no mention at JavaOne of the ECMA decision that it would elucidate the Standard for Java independent of Sun's guidelines. ECMA is the Standards body Sun turned to when it didn't get the control it wanted from ISO. Looks like ECMA isn't interested in rubber-stamping Sun's decisions, either.
Impressions: Mismanaged registration, long lines for the Palm Vs, but everybody upbeat. More exhibitors and attendees than ever, generally well-run technical sessions (the few that I attended: There were nearly 200 of them). Ran into Allen Holub, formerly Mr. C, in a hallway; he's a convert, now lecturing on Java threading. Yup, mainstream.
It wasn't long ago that I was attempting in this column to track all the new Java books that came out. That wouldn't make any sense today, but Addison-Wesley would like me to try, judging by the package the UPS truck dropped off last week: 10 new Java/Jini releases. JavaSpaces Principles, Patterns and Practice by Eric Freeman, Susanne Hupfer, and Ken Arnold (ISBN 0-201-30955-6) has a preface by David Gelernter, and it's worth reflecting on why Gelernter has a right to do the preface for this book on the Java/Jini approach to distributed computing.
When Bill Joy campaigned at Sun for distributed computing, it was Linda-style distributed computing that he was pushing, and Linda is the programming language in the context of which, in 1982, David Gelernter first described the idea of tuple spaces. JavaSpaces is the realization of that idea.
Tuple spaces don't require any direct communication among processes; as Gelernter puts it, for process alpha to get information to process beta, "the computer that housed alpha could be unplugged, taken apart, and recycled into beer cans and spare sand fifteen years before ground was broken for the office building that would eventually be torn down to make room for the factory in which beta's computer would be assembled." You'll notice that Gelernter is assuming that processes will be able to survive longer than the effective lifetime of some current storage media. I hope he's right.
I note from looking through Dustin Callaway's Inside Servlets (ISBN 0-201-37963-5) that people are still starting books with "Brief History of the Internet" sections. At least they're getting shorter: Callaway's is a half page. Soon they'll just say, "See RFC 1462." If you agree that Java servlets are the successor to CGI scripts, you may want to take a look at this book. If not, Addison- Wesley has another option for you: Robert Husted and J.J. Kuslich's Server-Side JavaScript (ISBN 0-201-43329-x). The authors have a lot of experience with JavaScript. They also have an anti-Microsoft bias, which you'll see in the book. I wonder about their claim that JavaScript is the most popular programming language in history.
For the last three years, there has been an ambitious project underway at Wolfram Research to give Mathematica the ability to handle gigabyte-sized datasets and beyond, especially for integer and floating-point numerical computations. Mathematica 4 is the first fruit of that effort. "An ordinary PC or Macintosh can now handle operations on large numerical matrices -- with a million or more elements -- and perform calculations with million-digit accuracy," according to the company. The Giganumerics project continues; it's not a 10,000-year view, but it's some kind of long-term project.
In keeping with Fielding's advice, I'm trying to think of my columnist job as being in charge of "a public ordinary, at which all persons are welcome for their money." Sliding his two bits across the bar, my frequent e-mail correspondent Felger Carbon writes to tell me of a plan to head off global warming or future ice ages. "Shall we compare the cost," he asks, "of providing enough additional heating to prevent an ice age to the cost of allowing an ice age to occur, for the world economy of 2100ad? 3000ad?"
Now there's an example of some long-term thinking. Check out Robert L. Forward's (yes, Forward!) temperature-control plan at http://www.whidbey.com/forward/ index.html.
DDJ