CL was in my opinion (and in the opinion a lot of Lisp hackers) a disaster in that respect. Before CL, Lisp had evolved rapidly by spawning new dialects. Once CL was established as the standard, this evolution practically stopped.
http://journal.dedasys.com/articles/2006/02/18/maximizers-sa...
My gut feeling is that there's a 'right' balance that's hard to find. If you just want to explore and be creative and push the limits, that kind of fragmentation is probably good, or at least not harmful. If you want to create something that people, say, base their businesses on, I think that it is not as positive. People get nervous about picking the 'right' one; beginners are especially confused, because they may not even have acquired the knowledge to decide what is 'right'. Lots of hacking goes into implementations, rather than libraries, and libraries may not work on all implementations, further exacerbating the 'which one' problem, which now involves looking at which libs run where.
Another part of the "problem" is that if you're aiming for "fun and experimentation", you are in a somewhat different (enviable?) position than most programming language creators: you've written about and popularized the language already, and are somewhat famous yourself, guaranteeing Arc a wide audience, something that is probably likely to attract more 'consumer' types than 'creators'. They're the ones who will be grumpy about not having something stable.
That sounds pretty good to me, actually.
I don't see why aiming for "fun and experimentation" would put me in a different position from other language designers. It's not like anyone does this for a living. I bet most of them started out working on their language in the spirit of intellectual adventure, then gradually got dragged into being motivated more by duty and eminence, till eventually like poor Guido their time was all taken up thinking about character sets. That sounds to me like something one ought to resist. Doesn't it to you?
Amusingly enough, you can see the pressure in action all over this thread, e.g.
http://news.ycombinator.com/edit?id=343663
I'm pretty good at resisting though ;)
In particular, you do not want everything in Arc to be an object. But the typical way other programming languages deal with Unicode nicely is to have different classes that implement a common string interface. So how can Arc approach the string encoding problem without making everything an object? I don't know, but I do think it's an important question.
I'm actually fascinated by these 'social science' (i.e. not really a science) aspects of programming languages. If you think about it, programming languages are all about people, how they interact with computers, and how they interact with one another to solve complex problems.
If I were an arclanguage community member reading this and seeing pg's responses, I would not feel better.
The whole situation sounds like a big mess especially considering lisp's history of "community fragmentation and trivially incompatible implementations."
I sincerely hope things get better.