That'd be great.
That seems like a good idea, since libiconv seems to be there on all *nix systems and there's a windows libiconv:
http://gnuwin32.sourceforge.net/packages/libiconv.htm, but I guess many string conversion stuff inside TeXmacs would've to be rewritten, and I have never looked into that. If everybody agrees and you feel like locating all the places where conversion is taking place then we could try it.
I think this was already fixed? Maybe it didn't work for CJK but it did for european characters? Well, I don't really remember and it certainly works now after applying your patch, so thanks!