Pavel,

> It's interesting that the keyboard and clipboard input on Windows doesn't
> correspond to any encoding (iso-8859-2, cp1250). Does the Windows VM convert
> this inputs internally (e.g. to MacRoman encoding)? I can correct a few
> important letters but not whole table...

  Windows VM does.  Don't modify the table for the sake of
consistency, but change ISO88592InputInterpreter (or
CP1250InputInterpreter)'s nextCharFrom:firstEvt:

to the following.

nextCharFrom: sensor firstEvt: evtBuf

	| keyValue |
	keyValue := evtBuf third.
	^ converter toSqueak: keyValue asCharacter squeakToIso.

The Windows VM passes the MacRoman-encoded (or Squeak-encoded) char to the
image, so first you convert it to original CP1250 value, and then
convert it to Unicode by CP1250TextConverter.

-- Yoshiki