I had refrained from criticizing Wave much, because I sincerely hoped to be proven wrong. On general principle, I support any attempt to rattle the cage of the entrenched patterns and technologies that inevitably constrain our communication even as they enable it. But such attempts are always Quixotic, as "entrenched" is more or less a synonym for "likely to defeat any attempts to supersede it." Now that this noble experiment has reached its end, I feel freer to point out the underlying realities that doomed it. They're predictable dangers for almost any attempt to be really innovative, and we ignore them at our peril:
The Siren of Coolness: Geeks are only human. Creative programmers and designers come up with all sorts of ideas. If the ideas are cool enough, they'll eventually get implemented. Coolness is hard to define, but it's basically geek appeal -- the likelihood that other geeky types will get excited about it. Unfortunately, coolness doesn't always correspond to usefulness or commercial potential.
In fact, the cooler a system is, the more likely it is to receive substantial levels of early stage funding, which means paradoxically that the coolest innovations are often the ones least likely to succeed. Less cool innovations face a more taxing set of evaluations, so those that make it through to funding tend to have real practical potential.
For five or six years I directed the research program for IBM's Lotus division, and it was a constant struggle to make sure that the enthusiasm of geeky executives didn't bypass the normal mechanisms by which research proposals and projects were evaluated. The people at the upper levels of companies like Google and IBM may be executives, but they're fundamentally geeks as well. The power of their positions doesn't automatically confer an immunity to technological enthusiasm.
Simply put: I suspect that to the decision makers at Google, the idea of Wave was too cool not to implement. It wouldn't surprise me if there was no market research at all until after the project was green-lighted. But the DeLorean was an insanely cool car, and how many of them do you see on the road?
The Chimera of Orthogonality: Imagine that you are designing a kitchen, in which you expect to have ten ingredients and ten tools for manipulating them. When the problem is viewed at that level of abstraction, you might naturally design the kitchen so that all ten tools have, to the greatest possible extent, equivalent levels of access to each of the ingredients. That's a kind of orthogonality: ensuring that all of the nouns and verbs in a system are equally interoperable.
This sounds perfectly reasonable, until you consider that two of the ingredients are potatoes and ice cream, while two of the tools are a baking thermometer and an ice cream scoop. Only by immersing yourself in the details of the way your system will be used in the real world can you design a system that people will find so useful and intuitive that they don't even think about the way it was designed. The ice cream scoop should be near the freezer and the baking thermometer near the oven, orthogonality be damned.
That's how I suspect one of Google Wave's most visible features went wrong. Sure it's cool to see an instant message being composed character by character, instead of waiting for the whole thing to come through. It was cool when it was done previously in research labs, and it was cool in Wave. What it wasn't, alas, was useful. For the sender, there's no margin for error if you take pride in your writing skills. The other person will see every typo, no matter how quickly you correct it, and will watch with interest as you agonize over the wording of a sensitive message. Meanwhile the recipient will end up spending more time looking at your message, as it slowly and hypnotically composes itself on his screen. It makes senders more self-conscious and vulnerable, while wasting recipients' time.
But boy, is it cool! And it's the natural result of orthogonally combining tools (instant mirroring of information) with objects (so-called instant messages). Sometimes, combining two great things yields lousy results, which is why you don't see too many combination barbeque and chainsaw sculpture competitions, or barbershop Beethoven concerts.
The Tar Pit of Complexity: One of the least commented on aspects of the ever-wider dissemination of computing technologies is the corresponding decline in the sophistication -- dare I say the technical intelligence -- of the average user. As an ever greater percentage of humanity is on the net, the average net user looks more and more like the average human being. It's often impolitic to admit it, but the truth is that anyone who grew up in an earlier era of computing is highly likely to overestimate his users. I'm not saying that users are idiots. But I'm not saying that they aren't, either. Increasingly, they cover the whole spectrum.
Given an ever more heterogeneous audience, software designers need to be ever more vigilant not to overwhelm the users. Learning new things is hard for many people, and most people are more comfortable sticking with the tried and true. Successful new technologies most often change only one or two things from what went before, thus minimizing the cognitive burden on users who try to adopt them.
Google Wave was full of good ideas. That was one of its biggest problems. Nearly everyone I talked to who tried it found it confusing, simply because it introduced too many new ideas at once.
The Underestimation of Inertia: The last fifty years or so have been amazing. Customs, conventions, and institutions have been replaced, one after another, in remarkably short time periods, through the advent of new technologies. Having grown up in this era, we tend to take such rapid and repeated change for granted, as if it's an easy thing to accomplish.
It's not. Consider email: For most forms of communication, email has largely displaced physical mail, the dominant technology for distant communication since the advent of writing thousands of years ago. That doesn't mean it's easy. First of all, it took forty years or so for email to evolve to that point, and for much of that time it was viewed as nothing more than a toy by most intelligent observers. It triumphed because it was, in the end, overwhelmingly superior to physical mail: it delivers messages in microseconds rather than days or weeks, its incremental cost is zero, and it enables levels of copying and mass mailing that have spawned a thousand new uses (and even a whole new criminal industry in the form of spam).
With such overwhelming advantages, it still took decades for email to displace what came before. It is sheer hubris -- albeit of a very understandable and common sort -- to begin any experiment with the expectation of that level of impact. Even in our age of miracle and wonder, radical changes remain rare. Our lives feel fast paced because change now happens over decades instead of centuries, but researchers and inventors routinely overestimate our receptivity to change. We are willing to change, but only if we are persuaded that there is a darned good reason, and that persuasion can take a long time no matter how good the reason.
Moreover, the higher the stakes, the greater the reason to change must be. There's a decent case to be made that the average home today would be better if it had DC wiring, rather than AC, all through the walls of the house. With a standardized voltage and connector, DC wall outlets would simply eliminate the need for all those annoying power converters that fill our houses today. There would be a small savings in manufacturing cost for every electronic device in the world, plus a small saving in energy efficiency. So, does that mean it's a good business plan to try to convert consumers from AC to DC? Of course not. Because of the huge entrenched installed base, the cost of the change dwarfs the benefits.
Similarly, to begin to displace email, or even to make substantial changes to the email paradigm, would require benefits comparable to those email had over physical mail. Good luck with that. Instead, successful innovation is likely to come with incremental evolution of email to include cool and useful new features, as we extended it in the 90's, with MIME, to include non-text and non-ASCII information, and as the IETF is extending it today to allow non-ASCII email addresses.
The good news is that Google seems to have learned this lesson. In talking about the demise of Wave, they say that pieces of Wave will live on, as they "extend the technology for use in other Google projects." Success in that endeavor won't be quite as flashy as what they hoped for with Wave -- but it's a lot more likely to actually happen. I wish them the best of luck.