On Sat, Aug 25, 2007 at 07:22:36PM +0200, Alain Bench wrote: > And if we had the magic universal isprint(), well... The practical > benefit to the auto-sensing guesswork would be much lower than expected. > "iso-8859-1:utf-8" would be possible, but unreliable (some UTF strings > happen to not contain 128-159 bytes, and would be wrongly sensed as L1). Try "utf-8:iso-8859-1" instead of "iso-8859-1:utf-8". For practical purposes, no L1 strings decode as UTF-8, imho. Kai