Concerning the encoding, one obvious consideration is whether existing
Squeak code can already rely on various characters being classified as
letters.  What does the Squeak distribution "know" about the top 128?

Character alphabet -> $a..$z only.
Character euro    This would need changing
Character nbsp    This would need changing
Character asLowercase  only recognises A..Z as upper case
Character asUppercase  only recognises a..z as lower case
Character isAlphanumeric  only recognises a..zA..Z
Character isLetter        only recognises a..zA..Z
Character isLowercase     only recognises a..z
Character isUppercase     only recognises A..Z
Character isSpecial       only accepts ASCII characters
Character isoToSqueak     would become ^self
String                    Case{Ins,S}ensitiveOrder only handle a..zA..Z

It's fair to say that while the Squeak character set is a subset of MacRoman,
most of the Squeak image hasn't actually caught on to this yet, so won't
notice if it's changed.

Naturally I have not been able to hunt down everything that might know about
various characters.