Dr. Dobb's Journal February 2000
In this month's walkabout we'll observe some of the smaller fauna of the field. These nature hikes are always primarily about identification, and in this one you'll get tips on how to identify web sites by their distinctive aromas, as well as how to distinguish one cyberlinguistic species from another, so you will never again confuse Rebol with Java. At this time of year there's even a chance that we'll catch a glimpse of an amateur mathematician. Let's hope for the best.
I never promised you a rose garden.
Such could be the response of the computer industry when and if computer users rise up in outrage against the stinking-up of the computer-using experience. When and if that happens. Of course DigiScents would prefer that you believe that its technology will only deliver sweet and welcome fragrances to your desktop, but don't count on it. DigiScents (http://www .digiscents.com/) is a Silicon Valley company that has figured out how to do what has only been the subject of jokes and hoaxes and fringe research in the past -- it has learned how to scent-enable your computer. Or your television set or your game machine, for that matter. The DigiScents device, which they claim they are seriously planning to market under the name iSmell, takes some digital data as input and produces as output a wide variety of selected smells. How wide a selection is a particularity that will have to await the actual release of the product, but a bunch, apparently.
As a video game machine peripheral iSmell could give a new dimension of reality to games, as a TV add-on it offers the promise of smell-o-vision (if that's a Lockheed trademark I apologize in advance), and as a computer peripheral it immediately suggests the possibility of scent-enabling e-commerce web sites. Just smell that cologne, that cognac, that Corinthian leather.
Joel Lloyd Bellenson, one of the company's founders, took some academic research on how odoriferous molecules trigger smell receptors in the brain, then apparently drew the shortest path between that research and a marketable product, and he and partner Dexster Smith followed that path. Their goal is to synthesize all smells from a relatively small palette, like the color mixing in an inkjet printer. The full scent palette may need to be a hundred times larger than red-green-blue or cyan-magenta-yellow-black, but the idea's the same. They have produced a box that is at least a first approximation to their goal: Send it the right signals and it emits the aroma of burnt wood, bananas, cheap perfume. The November 1999 issue of Wired magazine devoted a cover story to DigiScents.
I take quite seriously the money-making possibilities of this DigiScents technology. Obviously it's a boon to selling online products for which the aroma is an important feature. But it could be more broadly useful in more a subtle way: to stir emotions subliminally in product ads and political messages. Smell also offers another modality for storytelling in movies and games.
One application that will probably be promoted by someone is aromatherapy, the alternative medical practice of prescribing the sniffing of essential oils for health. I have real doubts about that one. If aromatherapy works at all it presumably depends on the effects of the specific oils and their chemical constituents. As I understand it, DigiScents' technology concentrates on triggering the right receptors in the nose, and may get its effects via quite different chemicals than aromatherapy traditionally uses. The opportunities for developers include hardware devices, smell cartridges, and smell-enriched content. DigiScents will likely try to license its technology broadly.
I'm also inclined to take the emotion and memory angles seriously. In the November 1999 issue of Scientific American you can read about Rachel S. Herz, a psychologist who is almost alone in doing serious research on the connection between smell and memory and between smell and emotion. The olfactory system is unique among the senses in being directly connected to the limbic system -- the amygdala, which is the emotional center in the brain, and the hippocampus, a memory center. The other senses are all mediated in their communication with the limbic system, but the olfactory system talks directly to memory and emotion; the limbic system actually grew out of the olfactory system, in evolutionary terms.
Taking off from this fact, Herz propounds the radical idea that emotion is essentially the same as scent, that it is just another, more abstract, expression of the same information.
On the other hand, she also says that her research shows that women consider scent the most important factor in mate selection, and men consider it the second most important factor. In a culture in which people take such pains to disguise their native scents, I find that result a little hard to believe, and that makes me wonder a little about all her work.
But just a little. The nose definitely has a rather direct path to our memories and our emotions, and we have never before had the ability to fully exploit scent as a communication medium. We don't know what the results might be.
It won't all be roses, I'm sure.
Unlike sight or sound, smells linger; and like sound, smells can't be easily contained. The developers have already run up against the permanence issue: Before you can appreciate the next smell, you somehow have to get rid of the previous one. Perfume vendors have come up with the trick of keeping bowls of coffee beans around to sniff. Apparently coffee sort of neutralizes the effect of perfumes. Whether that helps the users of the iSmell isn't clear, since they will be experiencing a wider range of smells that any perfume-counter sniffer would, but at least it gives me a Java reference for this item. An effective implementation apparently requires the use of Java beans. No word whether that has to be 100 percent pure Java, or if it could be any aroma starting with the letter "J."
But now the really bad news: In a world of computer viruses, e-mail scams, and spam attacks, we have to anticipate the offensive use of the technology. There will doubtless be stinkbomb e-mails, olfactorily sabotaged web sites, and smell viruses. Less threatening but more widespread will be the effect of bad taste. Think of the early days of desktop publishing, the days of ransom-note design. Some folks don't have good taste, and some folks have less sensitive noses than others, too. Porn sites will surely use scent to good advantage -- or bad, depending on your perspective. I'd just as soon drop that line of thought, except that wherever pornography rears its head, sticky legal issues ensue.
I wouldn't be surprised to see digital scents being banned in many workplaces, and in public places like libraries. This could create some interesting legal challenges. There is no specifically established right of free stench, akin to the right of free speech, but free speech has been interpreted to mean free expression, so an argument for freedom to smell could be made. That would immediately raise other questions: Are there smells that could be judged obscene, as opposed to merely suggestive, prurient, obnoxious, or offensive? Are there racist smells? Is it racist to suggest that there might be? This whole thing could be a political hot potato. Or make that onion.
Another legal issue that has some precedent: scents as intellectual property. Sun Microsystems might actually want to protect a particular coffee smell as the official 100 percent Java (tm) aroma, granting the right to use it only on products or web sites using approved Java technology.
I never told you, a couple of months ago when I was writing about the free Rebol programming language, that there was a commercial version in the works.
It's out now. Carl Sassenrath and his merry band of Rebollers have released REBOL/Command 1.0, an enhanced version with features designed for e-commerce application development, including support for ODBC and for calling third-party applications, and the ability to call platform-specific system or shell commands or DLLs from within REBOL scripts. As the reference to DLLs and shell scripts suggests, REBOL/Command will initially be available for Windows and popular UNIX versions, including Linux.
The original REBOL/Core is still available as a free download and still being enhanced and maintained. REBOL/Command is a particular package for a particular market. More such packages can be expected to come later. One news report on the announcement got a little carried away, characterizing this as the language that might succeed where Java has failed. Hmph. I'm a fan of Rebol, but that's a bit over the top. I won't itemize the many tools that Java has and Rebol lacks, since it could be argued that this is just a generation gap. Rebol is still very young. But even if you agree that Java has failed, a claim that would demand clarification in any case, I think the failure would be Sun's failure to carry through on the promise of "write once, run anywhere." That's mostly a failure of Sun itself to control Microsoft and to strike a workable balance between its desire to control the language and the concerns of standards organizations. It's not a technical issue. Rebol's approach to "write once, run anywhere" is for the Rebol staff to port the language to every platform they can think of. They've done a great job with REBOL/Core, but for evidence that the approach has its limits, see the platform requirements for that Rebol/Command release. The Mac isn't even on the to-do list, so far as I can tell, and some other platforms are likely to take a while.
I'm a fan of Rebol. It's an interesting, powerful, easy-to-learn language that manages to fit a significant bundle of capabilities into a remarkably small footprint. It makes it easy to write simple Internet applications -- easier, I think, than any other tool has managed. It may have a spectacular future, but it is not today a replacement for Java, and won't be any time soon.
I never claimed to be a mathematician, but I have represented myself as not being an enumerate boob, so it was inevitable that I'd get the math wrong. Anyway Lloyd Rice thinks I did. Lloyd wrote to point out that the future is not asymptotic. Most of the compelling evidence for the increasing pace of advancement in technology (see my recent columns on Stewart Brand's The Clock of the Long Now and James Gleick's Faster) is the kind of change regarded as generally following Moore's Law. But Moore's Law has no asymptote: If CPU performance doubles every 18 months, it just keeps on doubling. "If it's 100 MIPS this year, then it's 10 to the 32nd MIPS a century from now," Lloyd says. "No problem." Actually I think he's assuming a 12-month doubling, but the idea's right. Of course it won't happen; Moore's Law is already breaking down. Or is it? There is a growing sense that Moore's Law taps into some property of technology as a whole, and that the breakthroughs will come when the limits of particular technologies are reached. This sort of naive faith in technology seems -- well, naive, but it also seems entirely consistent with recent experience.
Still, even if technological trends continue to ramp up exponentially, that doesn't lead to an asymptote. But the argument is that there's more afoot than Moore's Law, or rather, than whatever forces lie behind the useful observational extrapolation that is Moore's Law. Gleick and Brand report on the views of assorted more or less astute observers, many of whom see progress going kerflooey in a few years. There is some sloppiness in the way some of these people talk about the data, it's true. But the notion of an asymptote, a singularity, comes from plotting lots of trends, I think, and observing where the best-fitting curve seems to go. The best-fitting curve isn't, apparently, a Moore's Law-obeying exponential curve. At some time in this century the curve really does seem to go vertical. Progress becomes infinite. So say the curves.
What that means is hard to say. I think it's safe to say it doesn't mean processors will have infinite speed in this century. I suspect it just means that we are reaching the end of our ability to predict the future. But that's plenty for it to mean.
I never sold you an Edsel, but I did dedicate some precious DDJ pages a couple of years ago to sharing my experiences in becoming a Newton software developer. Who can forget my mushroom identifier program? If I had only waited until DigiScents came out with their iSmell device, I could have incorporated the all-important dimension of smell into the program.
Of course, iSmell doesn't work with a Newton MessagePad, but these days who does? Not I. I have joined the rest of the human race, leaving the company of fanatics. Well, leaving that particular group of fanatics anyway. I don't give up easily, as the abused deceased quadrupeds and currency-stuffed rodent holes around here attest, but I do eventually get the picture.
So I'm not actually going to exhort you to become a Visor Springboard developer in your spare time. In fact, maybe you shouldn't. But you might be tempted. Let's consider the pros and cons. The Palm handheld device platform has been a runaway success, and a lot of people have written little programs for the Palm. Some people have made money, but there are complaints that Palm and other companies have been giving away Palm software so freely that they are making it hard to make a buck selling the stuff. Now Palm is espousing some sort of services model, which doesn't seem particularly helpful to third-party developers, either.
When the Palm founders went off to start their own company, licensing the Palm OS and coming out with a cheaper, better device, it looked like here was the boffo platform, especially with those add-in modules. But the lower price of the Visor can reasonably be expected to lower consumer expectations about what they should be asked to pay for software and peripherals, once again making it tough to make a buck doing third-party products.
Still, a tiny piece of a big pie is sweet. At last fall's Comdex trade show, the Handspring's booth was swamped. Voices in the crowd were heard comparing Handspring's Visor with the Palm device (a Wired news reporter caught this: "This is thinner, and you can upgrade and get more memory without having to open it... the software is easily transferable; the expansion slot is the thing, plus, it's a nice funky color"), and with Windows CE devices ("It takes more power and guts to support CE...it will put Windows CE out of business").
The development opportunity is in those Springboard modules. Visor's Springboard feature is the first real plug-and-play capability for handhelds. Just plug a device into the back and the device's capabilities are immediately displayed on the Visor's screen. No drivers to install; plug in a cellular phone device and Visor becomes a cellular phone, plug in a GPS module and the Visor is a GPS device. The developer's kit is freely downloadable from the Visor site (http://www.handspring.com/), the Visor folks seem developer friendly, and the development process is simple. Some people are going to make a lot of money developing Visor modules. So it seems to me.
Reports are that Palm hasn't been quite so helpful to developers. Nevertheless, there is a PalmOS emulator in Java that I feel I must mention. You can read about it at http://www.javaworld.com/javaworld/ jw-11-1999/jw-11-device.html. Whether it's Visor, Palm, or Windows CE, I see a lot of software out there now for handhelds. Hmm, there's a fantastic opportunity. I could port it all to the Newton! And publish the code here! Only kidding.
DDJ