I disagree with this.
There are things you can (are allowed to) do, things you know you can't do, and things that are inconceivable / impossible.
His position is that #2 is evil (because knowing you could fix that printer driver if only you were allowed to is fucking annoying), and there is a moral duty to minimize it.
I think that #1 is good and should be maximized, and there is very little practical difference between #2 and #3. Minimizing #2 at the expense of also reducing #1 is bad and harmful.
In practice, there is a flow of things from #3 to #2 to #1. Things are moved out of the impossible with funds from paying customers, and then made open by people re-implementing them once the hard exploratory part is done.
There is also a smaller flow of things directly from #3 to #1. Some research is publicly funded or due to individual curiosity, instead of corporate internal. Some kinds of products can be funded from support instead of initial sales.
Destroying category #2 would choke off the first, much larger, flow. The second would not be able to make up the difference. This result would be good according to Stallman's ethics / priorities. I think it would be bad.