Ken Arnold's Blog

Business Archives

Patentably idiotic

The Kodak v. Sun suit has gone against Sun. This is hard evidence that the software patent system is deeply broken. I know this isn't news; you probably already knew.

One approach is to think that software patents are just plain wrong. Maybe so, but this isn't obvious to me. Patents have protected other technologies, and they might be able to handle software. Software patents as currently done are broken, dangerous, and frankly f'ing insanse, but maybe it's fixable. Or at least we can lower the temperature of this hot tub in Hell.

Any mitigation must start with the basic problem of knowledge: Almost nobody with actual authority in software patents is competent to understand software. Patent examiners (I'm sorry to say) are not our best and brightest. The Patent Office doesn't pay enough to attract or keep top level folks. Juries (like the one in this case) are picked from the normal jury pool. Imagine trying to explain the difference between (say) virtual memory with paging vs. an on-chip CPU cache to mother. And then remember that your mother is smart enough to avoid jury duty. Not to mention that the judge is a lawyer, most of whom have only the most basic scientific training.

In short, everybody who has the power to make judgements is almost certainly unqualified to do so.

My favorite solution would be to fire all the examiners and replace them with 25 actual experts, and let them grant (say) 30 patents a year (with maybe a unanimous vote for adding more).

But we could also require that the "jury of peers" be peers of computer people. Those of us who are computer folks could join that jury pool and be exempt from other jury duty. Then maybe the jury deciding this kind of stuff might have a chance at having a clue. This would be especially good because patents are only supposed to cover thing not obvious to your average pracitioner in the field. Who better to judge besides a jury practiioners in the field?

We could also make software patent grants provisional for 6 months. They get published and the outside world has a chance to throw prior art at it to see if it sticks. A lot of companies would be highly motivated to try knocking patents down before they cause any damage. It would be a way of drafting highly competent technical folks from these companies into the process of weeding out the idiocies.

These are just some ideas, but it's a pretty depressing situation and we oughta do something...

Knowing What's Where

So far today the most interesting talk I've been to is the one on RFID techonology by Sun. They've built it on Jini and Rio, which itself is built on Jini, and they have an impresive platform for dong stuff with RFID applications. The match is good because these systems require customization and smarts at the edge of the network, where the data is gathered, because otherwise too much data would be sent back to the central processing systems. Jini is built on the idea of moving behavior to where it's needed. They also have to deal with failures all the time, so having something that is low maintenance to install and heal is critical.

What's going on with RFID itself is also interesting. We're getting to the point where tracking objects -- real objects -- will be a major focus of computing. When you can know where every piece of paper is in your files we will have a new scale of relationship modeling. (Well, new to most -- in the past anyone dealing with this much specific information was using large-scale systems that were very hard to build.) I think there will be a lot of interesting work in figuring out how to correlate all the information in a timely fashion.

Tomorrow's candidate for "Most Likely to be Wow" is the Orbitz talk. Ever thing you book on Orbitz runs on a Jini backplane.

This work is licensed under a Creative Commons License.