Youtube is further proof that worse is better.
Lisp is a brilliant failure, by a bipolar student.
To paraphrase Tony Hoare, premature commercialization is the root of all evil.
If you have something that depends on a network effect for success, and you decide to charge everyone money upfront, you won't get the uptake necessary for platform success. You won't create the large-scale network effect which would create your future market of customers.
So if you come up with some nifty new programming language, and want everyone to use it, don't immediately go start a company and try to sell the compiler. You have to give it away. Maybe you can bake some kind of upsell into the thing, maybe not. Same goes for operating systems, social networking platforms, whatever.
Another way of saying this is that you should trade revenue to get market share.
It's interesting that the computing platforms we use today seem to have come out of either academic distribution or a quasi hippy Berkeley culture. "Code should be free". If you go all capitalist on your platform innovation too early, it goes nowhere.
I was trying to research Dick Pick, of Pick Systems, the guy who built the database language, to find material for this post, but Google kept messing up my queries. You can't search on Dick Pick. You get all these hits for "Dick's pick's". I sat in a meeting once where Larry and Sergey explained why stemming was bad. They were right. They should have stuck with it.