Quotes of the week

Again, it's a trade-off of short term versus long-term viability. BSD folks are happy anyone is using their stuff, which is fine, but I don't think it's worth it if the cost is getting code contributions back by the community.

It is in the long term that GPL helps. The example of proprietary binaries is a good one. Over time we see more and more companies moving away from this approach and moving towards actually releasing source code. The pain in the ass to users of having to target specific kernel versions, or not being able to use custom builds, and so on, is just not worth it. Network vendors get it, disk vendors get it, and even graphics vendors are starting to get it (Intel has for years, AMD is finally coming around, and... well, NVidia... never mind).

Sure a startup can whip up a proprietary kernel module and sell the fuck out of it until they can get bought by Google or Cisco or whatever their business plan is, but in the long run it's a costly solution.

Still, Apache projects like Hadhoop seem to attract external contributions despite their permissive license. Isn't that quite counter intuitive to the rhetoric of copyleft proponents?

> If someone puts a sign up and says "free donuts, donate what you like" and you just take the books and don't leave any money... well, it's not very social but you're not doing anything wrong. But this is why I think that copyleft licenses are necessary in the long term - otherwise companies will just [ruin] the commons.

First, taking the books would be theft, since it was the *donuts* that were free... ;)

But more importantly, we're talking about an infinite supply of donuts here. Whether you choose to publish your changes or keep them secret, making use of code (BSD-licensed or otherwise) does not prevent others from doing the same. One could argue that contributing back would make everyone better off, but you certainly aren't making things any *worse*. Not contributing back may not be "social", but it isn't anti-social either. If using the BSD-licensed code at least means that your customers get a better product than if you'd decided to roll your own ad-hoc version, some good has come out of the arrangement. The commons isn't enhanced, but neither is it diminished.

If the license were copyleft instead, it *might* mean that the modifications are made available to the users—or it might just mean that the product doesn't exist at all, or that it exists but uses some idiosyncratic proprietary from-scratch implementation riddled with bugs and incompatibilities, either of which would be worse for all concerned than a version based on BSD-licensed code with some proprietary modifications.

To use the Netscaler example, are BSD developers writing complex, stateful, deep packet inspection? If so, then why are they spending their time on something so specialized? If not, then how can there be any competition?

(yes, of course bpf is getting very impressive, and on modern hardware might even keep up with last decade's Netscaler, so maybe one could claim there's starting to be some overlap... but that's just avoiding the question.)