Wouldn't it be sensible to instead contribute to the compilation framework of PLT Scheme instead since Arc is hosted on PLT scheme, or is there a solid reason to not do that?

This could be a step towards self-hosting for Arc.

I think that thats a bit premature, given how immature the language specification apparently is, and anyways there are few implementations for languages which have the syntax and runtime error handling capabilities of plt scheme has, and the core language is rich enough that in terms of interpreter / compiler performance you won't get much better without sacrificing some nice bits of lispiness...

on the other hand, knowing all the weaknesses that pop up in the language when used in a standard compilation setup is always good

Um.

Why not use Chicken: http://www.call-with-current-continuation.org/

or Bigloo: http://www-sop.inria.fr/mimosa/fp/Bigloo/

or Stalin: http://cobweb.ecn.purdue.edu/~qobi/software.html

or Gambit: http://dynamo.iro.umontreal.ca/~gambit/wiki/index.php/Main_P...

And for those of you who think that this would be cool if it was mature and stable, GIYF.

This would be neat if it was complete, but it's far from being usable.

Yes, it's veeeery far from being usable. But, hey, it's only been started 6 or 7 days ago...

Please update us when you have something reasonably stable.

And note the first release has no problem with Unicode :)