I've made three very slight changes to AMC-ACE-Z, one to fix a small bug, and two to make it very slightly simpler and a little more flexible, with virtually no effect on the encoding length. The new spec is available at: http://www.cs.berkeley.edu/~amc/charset/ I won't be able to submit it as an internet-draft until after the next IETF meeting. The rest of this message describes the changes. First change: Formerly, LDH characters were encoded literally and non-LDH ASCII characters were forbidden. Now all ASCII characters are encoded literally. This has no consequence for LDH names, but is reasonable behavior for less restrictive names (like the names of SRV records). (And it matches the behavior of UTF-8.) Second change: The initial value of n is now 0x80 instead of 0xA1 (this is an incompatible change). With these two changes, the encoder no longer needs to check for invalid input, because there is no such thing anymore; AMC-ACE-Z can now encode any Unicode string as an ASCII string. The ASCII string will contain non-LDH ASCII characters only if the Unicode string does. Third change (bug fix): The decoder had neglected to verify that the input contains only ASCII characters. AMC
