I would not revise FCD itself. For a number of processes, it is sufficient as is. For collation it's not.
About the Tibetan precomposed vowels: For the LDML spec, I submitted a CLDR ticket this morning: http://unicode.org/cldr/trac/ticket/5667 For UTS #10 section 6.5, I just now submitted an error report on unicode.org . About the other issues: I think that's up to implementations. If an implementation wants to avoid the NFD step then it needs to build more data or add checks for problematic input, or both. The spec should not go into that detail. For ICU, we have a statement in the User Guide that overlaps between contractions and decomposition mappings are not supported, and we have a ticket for trying to fix this by building more data. markus
