He argues that some features - e.g., dependent typing (http://axisofeval.blogspot.com/2010/10/notes-on-dependent-ty...) - won't emerge from Lisp, because they approach the problem from the wrong angle; no safety towards more safe vs. Haskell (and others') reverse approach.
Good point - I forgot to consider this approach. I'll have come back to it at another point.
Generally though, I think Haskell's approach of re-adding freedoms is more fruitful than trying to add bondage to Lisp. And "the Haskell movement" certainly has a different kind of manpower behind it.
There are probably hundreds, if not thousands of researchers, who use Haskell (and related PLs) as their vehicle for type system research. People working on "gradual" type systems like Typed Racket are probably around a dozen, which means that progress in Haskell happens much faster and with more breadth.
While i think you make valid points, i do not agree with the conclusion.
First, i don't think associating innovativity with numbers of researchers is a safe predicate since it's clearly not linear.
Second, the fact that haskell is a testbed for researchers doesn't make it the ideal language for "real world programming" in my opinion. It could be that it is, but again the correlation is less than clear.
And third, and that's the most important point in my opinion, the ideas developped in haskell are fully available to other language implementors once they exist, and the experience of researchers is available. It means that a language like typed racket or any other could possibly implement the best/most usefull of those ideas in much less time than was needed to first find them.
When my post started a heated conversation in ru_lisp LiveJournal community, someone told there about his personal experience with Qi. He managed to convince Qi that 42 is of type String.
