ESR has a point in that the GPL is less needed to prevent forks than it may have been because forking hardly pays anymore or only pays in the short term, because you simply can't keep up in the long term.
However, that only applies in the situation where your contribution is small, and the work you're adding to is relatively large, such as Brand X GCC or Brand X Linux. And even there, considering how many people in kernel development track a moving baseline while maintaining their own patch sets intended for eventual inclusion in the baseline, the strategy of maintaining a proprietary addition does not seem as infeasible as ESR makes it sound.
You don't have to duplicate all future work by the internet to stay current. You only have to keep merging your own stuff. In short: the penalty for leaving the main branch, the Open one, isn't nearly as high as he implies.
A situation where it does not apply at all is where a large proprietary work successively absorbs small but innovative open source works.
ESR completely ignores this scenario. The small work does need to be tracked by the proprietary company, but such tracking is already a fact of life in large development projects composed of many smaller subprojects. An open source project would just be an unofficial subproject or vendor component that would have to be tracked and re-merged whenever new features are released that are attractive.
Without the GPL to protect the smaller works, they'll simply help the big proprietary projects to stay up to date.
Even if, /if/ our production model is so superior that we can win even while contributing to the other model for free, I personally feel that companies that don't pay me, have no right to impose any more licensing restrictions on /users of my software/ than I would impose.
Hence my choice for the GPL for everything I release publically. If a company wants to restrict indirect users of my work, they'll have to pay me for a different license.
That provides a source of income that I would lack if I would only make money on services. If /everyone/ was in that boat, I would not care because I get other people's software in return for my software, allowing me to provide the same services as everyone else, and compete. But as long as proprietary software companies want access to my work, without giving me access to theirs, I need to be compensated in a different way.
ESR ignores that mechanism too. I don't know how important it is to the overall picture, but it /is/ important to me.
When two methods of production provide the same total output or arrive at the same balance overall, one may have much larger variation or a much larger internal tensions in the balance than the other.
Same total output is not the only metric to judge a system.
If you can produce as much with slaves as with people with rights, then that only says that the systems are equal in terms of output, not that the systems are equal.