I'm very much in favour of the first one. Simple things like that can have a big impact on how people perceive a language. I even began writing an IDE in Arc as a staring point, but it got reprioritised.
Even better though would be if the Lisp community put down its stupid in-fighting and built a language-neutral easy-install environment; something that could edit and run Scheme, CL and any other dialect you might want to write. Imagine building Arc on an environment that's actually designed to have Lisps implemented on it, rather than trying to fake CL style lists in MzScheme and all that nonsense. Imagine if you could write back-ends to different architectures, making Rainbow and arc2c unnecessary.
That would be cool. As far as I know, s-exps are s-exps no matter which lisp they're written in. The general environment might need a good way of defining syntax so that it could be more helpful than just a text editor with REPL, but that sounds like something lisp should be capable of doing pretty well.
The only problem is that everyone would argue over which language to implement it in. How about we just start in arc, and let them all join us later?
Actually, it sounds like an extended DrScheme, which already supports several different languages of scheme in the editor. If the interface was improved a bit, and it was extended to handle CL and Arc, then it should suit admirably. Unfortunately, I don't know how the language selection feature works, so I don't even know if that's possible.
Arc is probably a good language to implement it in, because it has such a small core and is so flexible. I also think Arc's users are a bit more experimental and ambitious than most Lisp users. Consider the fact that Arc will probably have a shared-nothing garbage collector before any other dialect!
