On Friday, 3 June 2016 at 12:04:39 UTC, Chris wrote: > I do exactly this. Validate and normalize. And once you've done this, auto decoding is useless because the same character has the same representation anyway.