This is complicated. Unicode's case-folding algorithms provide defaults, but, as we discuss later, there are at least some locale-specific foldings (which, I suppose, one could consider part of Unicode's case-folds......)

What about something like this:

By default, string searching SHOULD be case-insensitive. Implementations SHOULD use Unicode's default case-folding algorithms, with locale-specific tailorings applied when the language context is known.

I think that's a good overall statement, but it's making the advisement do too many thing. Perhaps break it up like:

• Allow users to switch between case-insensitive and case-sensitive searching
• Make string searching case-insensitive by default
  • Use Unicode's default "full" case-fold for case-insensitive searching by default (where full is defined as common plus full mappings)
  • Optionally apply locale-specific tailorings of Unicode's "full" case-fold for case-insensitive searching when the language is known.
  • Use Turkic case folding where needed (more work needed for this one)