I found the question at the end an interesting one (paraphrased): "would you be happy if people wrote websites to sell dog food in haskell?"
I was a little disappointed in SPJ's response, actually.
What would make it a bad language for such an application? What would make it a good language? Would the answer change if you instead wanted to create a framework for building thousands of web-store applications that were all similar but a little different?
As I understood the question, it wasn't really about Haskell's fitness for that particular purpose but rather a bit of a jab at its theoretical, researchy nature. I thought it basically meant "How would you feel about your baby being used in a completely quotidian manner?".
I have been trying to learn haskell for a while, and I like it quite a lot. I found myself thinking things like "what kind of amazing software could I build in haskell that I couldn't with any other language". I realized that was the wrong way to think about it, but perhaps there's something more there. After such an investment in learning, maybe writing a simple web store seems like an under-use of the language.
I wonder what things would be like if people did write ordinary applications in haskell.
Well, people do write ordinary applications in Haskell (it's not just banks :)). And the result is surprisingly stable and short programs. A perfect example here is XMonad--it's a fairly popular tiling window manager with a good set of features and is something like 2500 lines long (just using wc -l to count, so don't trust this number :)).
Additionally, Haskell is actually pretty well suited for web programming as well. There are several frameworks which make writing server-side code easy and have very good performance; the two most popular ones seems to be Snap[1] and Yesod[2]. The combination of expresiveness and performance Haskell offers server-side is very good--C++ may be faster but is uglier, Ruby and friends may be as expressive but are slower.
I wasn't talking about the suitability at all. I'm still learning so I can't comment on that, but it seems fine so far.
Maybe another way of thinking about it is: I wonder what it would be like when haskell gets to the point that an application written in haskell is distinct or ordinary because of what it does, not because it's written in haskell.
And that kind of goes back to the question during the talk: there's nothing distinctive about a website that sells dog food, so if people start choosing haskell for that kind of thing, that would be a turning point.
I was a little disappointed in SPJ's response, actually.
What would make it a bad language for such an application? What would make it a good language? Would the answer change if you instead wanted to create a framework for building thousands of web-store applications that were all similar but a little different?