How could Squeak's support for Unicode be improved?
more unicode to character code translators.
review existing code and clean it up, some classes have methods that have different code to do same thing.
check the code for correctness, and fix it (example: the unicodeToMacRoman actually does unicode to latin1).
Improve performance of the translators: as an example the UTF8 translator creates a write stream, but then actually writes out bytes to your file system one byte at a time as it considers if the unicode byte needs to be made into a utf8 sequence. Needless to say if you have a multiple MB text file this suddenly takes millions of file I/O requests to the operating system which is really slow.
Earlier notes 2006
Some years ago, Yoshiki Ohshima had a japanese version of Squeak, and recently he presented a proposal for Unicode support and a first implementation. Details are on this page: Multilingual Squeak
It would be a good next step to select the best from both proposals and to create an unified implementation proposal. Selection of a common implementation policy should also examine the question of possible or desireable changes in the VM. A lot of problems are still not solved. For example, we do not currently have the possibility to use one of the IME-s (Input Method Editors) to enter text in Unicode.
