Our ZDNet sister site in the US has been hosting a discussion about what 'open' actually means. This discussion was initiated by David Berlind, who's been gnawing away at the issue for a while.
One of the most interesting ideas came from Bob Sutor at IBM, who suggested a scale of openness. This could range from "I retain full rights over what I sell you, and you may do nothing except the single task for which it is licenced" to "Do what you will with this". Something like BSD would come towards the latter end, with the GPL slightly less open — you're not allowed to make GPL IP closed — and our favourite proprietary operating systems well towards the former. Microsoft's concept of open formats, where you have to have a licence with the company before you can use them, isn't very open. I couldn't implement a system and then give someone else the code to use.
But this demonstrates why "open" is such a bad term. Like free, it is an absolute term being asked to contain a multiplicity of nuance. A much better axis would be limited to unlimited. An unlimited licence to all information about a product is unambiguous. A licence with one degree of limitation — that you must impose it on all derivatives — covers GPL. One with two degrees of limitation, that you cannot pass on the licence and the licensor maintains control, covers Microsoft's open standards. One that forbids access to source code, forbids reproduction, and limits modes of use would be three degrees — and this moves into the territory of copy-restricted media and the systems that use them.
The idea's not perfect; it is impossible to accurately categorise the drifts of dense legalese that characterise the average end user licence agreement in such crude ways. But as a rule of thumb, it has the advantage that it prevents the abuse of 'open' by forcing the recognition that there are limitations. It also aids functional comparisons: BSD is zero-L, GPL is one-L, WS-* is two-L, Windows is three-L. There might also be an interesting correspondence between the physical length of licence agreement and L-ness, which would be a pointer towards an objective categorisation system.
Simple minded? Undoubtedly, but useful nonetheless. Will there be arguments about the definitions? I hope so — even if that's the only thing that happens, it will force people to confront the fact that 'open' has been roundly abused for years. The currency is debased. Are there important areas that aren't addressed by this idea? Certainly: pricing, compliance, the difference between standards, systems and implementations, are all dimensions related to the idea of control inherent in licensing and which aren't easily compressed into a simple one-dimensional scale. An accurate comparison of the limitations of different sorts of licence would require a complex matrix of concepts: you can compare apples and oranges, but you need to get into biochemistry.
We don't need to get into that level of detail to define the important stuff. Limitations in licences matter more as the licensed products matter more. If I'm buying software to control a model railway layout, I probably don't care if it's locked down tighter than J D Salinger's Christmas card list. If I'm choosing a security system for general purpose e-commerce, then I'm going to care enormously if it hands control to a third party. There are significant general principles here already touched on by antitrust legislation: you don't get to win overall control of a market without acquiring new responsibilities for acceptable behaviour.
It's a shame that such ideas need to be codified. For most of the history of innovation, a balance has been struck between the importance of an idea and the protection afforded to its creator. Corporations have learned they can game the copyright, licensing and patent systems for commercial gain, thus upsetting that balance. Until systematic reform is effected, the least we can do is make the mechanisms of those games plain and encourage a proper appreciation of the implications of our choices.
