If I recall correctly (it's been a while) a good way is to use Ada fixed-length strings wherever you can, and use something like Dmitry Kazakov's UTF-8 library in other places. People often use unbounded strings in places they don't need to instead of declaring a fixed-length string when it's needed.
Nice to see somebody's still using Ada. I gave up on it for new projects because of the lack of infrastructure and 3rd party libraries. It seems that half of comp.lang.ada consists of professionals from the aerospace and embedded systems industry and the other half are crackpots who want to create a Caesar cipher plus spam. It always looked like a weird community to me (but that has indirectly to do with how Adacore conduct their business).
Nice to see somebody's still using Ada. I gave up on it for new projects because of the lack of infrastructure and 3rd party libraries. It seems that half of comp.lang.ada consists of professionals from the aerospace and embedded systems industry and the other half are crackpots who want to create a Caesar cipher plus spam. It always looked like a weird community to me (but that has indirectly to do with how Adacore conduct their business).