> The best
> of our graduates know Haskell and especially its advanced features and
> they are hired with the greatest pleasure. Although at work, they
> develop most programs with more mainstream languages, their employers
> don't want to miss the high level of abstraction the Haskellers have
> mastered. The next big step is dependent typing, say in Agda or Idris
> (bootstrapped with Chez Scheme, by the way!).

Companies like to hire people who have wrestled with harder problems
than they are likely to encounter at work. Same reason math/physics is
valued for tech jobs.

The real value to society from programming languages is to make code
shorter while maintaining other desirable qualities (efficiency, safety,
simplicity and ease of understanding of the tools and the product). Is a
good Haskell or Idris program shorter than a good Scheme program? GHC is
a giant program and Haskell performance is hard to reason about. The
market has consistently voted against Idris/Adga-level complexity in the
name of safety. (ATS is an earlier contender in the same vein, with no
users.) The APL family has a higher level of abstraction than Haskell
with shorter programs using fewer concepts, and orders of magnitude
lighter tooling. Things like Mathematica and Torch are probably denser
as well, though not necessarily lighter than GHC :)

> And Mozilla had a reason
> why they pushed Rust. If there was an ISO standard and an independent
> implementation (say in GCC), I don't see a reason why to develop new
> programs in C++ anymore (ignoring legacy restrictions).

Yes, Rust is probably a big improvement over C++. But they need to
re-implement Firefox in Rust for the final verdict. C/C++ are the
hardest languages to displace. That must mean they hit a difficult sweet
spot.