Frank's Technotes

A new way of doing things.

I just read The Internet Is Broken at Technology Review. In it, David Clark at MIT asserts that the Internet needs to be, not fixed, but replaced with something better, something designed from the start to deal with the problems and the sheer size of the current Internet.

I think he's right. In fact, I've been thinking along those lines for some time, only I've had something a bit more radical in mind.

From 1994 to 2000 I worked with Bruce Walker and others through four different companies on single-system-image clustering software. Our team worked on what eventually became OpenSSI. The system embodied a lot of very good ideas, but unfortunately was a victim of indifference and the "good is the enemy of the best" phenomenon. It's still alive, after a fashion, but it never really took off. Personally, I think that our biggest mistake was in not thinking big enough.

Imagine a world in which your PDA, laptop computer, watch, DVR and desktop computer all talk to each other automatically, sharing not only data but computing cycles, memory, disk space and network connectivity. One in which the environment you use at home is exactly the same one that you use at work or in your car, right down to the fonts and wallpaper (not to mention the applications). One in which you never have to worry about a virus or trojan horse, since the authentication and authorization mechanisms necessary to prevent such things are built into the fabric of the system.

Imagine being notified on your laptop that your car needs gas or that it's time for its regular maintenance since it has passed 3500 miles since the last one. Automatically, because agents that run all the time have contacted the car and gotten its state without your intervention.

Imagine being in the middle of some particularly important and knotty problem at work when it's time to go home, then going home and picking up exactly where you left off, without doing anything special at all.

Imagine never, ever being offline because the mesh that your personal devices are a part of maintains its connectivity constantly, despite failure of individual components. Imagine building a sensor network or a compute cluster just by getting the appropriate hardware and doing some minor configuration.

Some of this is already happening, with meshes, grid computing, metropolitan-scale wireless networks, sensor nets and mobile ad-hoc networking. Some of it is a bit science-fictional. It's all possible, though, even practical. All of the problems have been or are being solved. The challenge is in designing the system to be robust enough to be useful and then in implementing that design.

And I have a few ideas about how to do it.