Re: Sprouting like mushrooms (or are they toadstools?)
How many languages that appeared 10-15 years ago are still in demand?
For the past 10 years: Rust (2010), Dart (2011), Kotlin (2011), TypeScript (2012), Julia (2012), Swift (2014).
I'm not claiming those are all good languages, but they're all "still in demand" by any reasonable metric. They're all being used for production applications, they're all still present in the trade media and in various surveys, they all still have their proponents and backers.
And I don't think that's a very useful metric anyway. There may not be much demand for AUTOCODER; that doesn't mean it wasn't important. You don't see a lot of new ALGOL projects - that doesn't mean ALGOL wasn't hugely influential. Pascal1 has largely fallen out of favor (pace Stob and other Delphi fans), but it left its mark, too. ML was never much for production apps, but its descendants OCaml, Haskell, and F# - even while they remain niche languages - have had a significant impact. Erlang has never been as popular as it deserves to be, but people keep reinventing it, so it must have done something right.
On the other hand, Fortran, COBOL, PL/I, etc., not to mention various assembly languages, may not be sexy, but they have a hell of an existing code base and there's still plenty of fresh development in those "legacy" languages.
And, finally, why not develop new languages? Kotlin and other JVM languages pushed Java to improve in its expressibility and syntactic sugar. If people are going to continue to insist on developing huge applications in ECMAScript-based languages, then yes, please, let's have some with a bit of type safety and other improvements on the base language.2 Whenever I'm writing something in Managed OO COBOL I'm glad to have generics and type inference and anonymous methods with proper closures - even if I don't need them for whatever I'm doing at the moment.
Language development gives us better languages.
1The programming language, not the mathematician/philosopher, or the Reg regular commentator.
2Yes, I know you can do purely functional programming in Javascript with proper algebraic structures and monads3 to defer side effects. And that's great, since you can then do all sorts of handy reasoning and manual or automated proofs of correctness about the vast majority of your code base. But clearly only a vanishingly small fraction of Javascript programmers are willing to learn how to do this.
3Look, we've explained this already. A monad is a monoid in the category of endofunctors, like a semicolon with side effects. It's so obvious.