> Shouldn't encoding: be encodingNamed: ?  There will likely be
> TextEncoding objects in the future, if there aren't already, 
> for various reasons.

Well the 'encoding' is the name of the thing the 'encoder' is the guy
handling it. I don't know yet if we will ever have the need to have an
explicit encoding object (right now I don't see any) but even if we do,
there's a simple way around the problem merely by relying on polymorphism,
e.g., a message like "asTextEncoding" implemented both in Symbol and
TextEncoding would solve the problem.

> The discussion of line-ending translation confuses me.  If you set
> #none, will the behavior of *input* also be changed?  If you set #lf,
> will that affect the input translation or just the output?  I believe
> there should be two settings, really: input and output.  (and in that
> case, #none is no longer needed, because it is simply #cr).  Also,
> something like #any may be a good option for the input direction.
> 
> Or is input always #any ?  If so that clarifies it all for me, except
> for the comment about #none.

#none is like StandardFileStream. Any of the others is like CrLfFileStream.
Hm... thinking about this - might it be better to call it "raw" instead of
none?

Cheers,
  - Andreas