> >Why is Canvas a subclass of NullEncoder and FlattenEncoder, lately? 
>
> I think you gave the right answer already and also suggested the right  
> refactoring to get rid of that inheritance dependency (which btw,  
doesn't
> allow to remove postscript printing completely).

Encoders actually have nothing to do with Postscript, except that  
the Postscript support is also built on top of them, so there is no  
dependence  'Canvas --- depends-on---> Postscript support', which  
would clearly be wrong.

The naming of the category they're in is misleading (and not mine).   
Anyway, as anyone who actually had a look at the code would figure  
out instantly, Encoders are an extremely light-weight abstraction  
that factors out some patterns that commonly occur.  They have been  
built very carefully not to include any extra 'fluff', which I would  
have thought is a *good* thing, but which seems to cause trouble with  
no end in sight, because everybody seems to just jump at  
conclusions.

Sigh,

Marcel