der Mouse wrote:
(a) Take off your character-string blinders for a moment! Your statement is true only if both octet strings represent character strings, and it's equality of the character strings that you really care about.
...which they do, for pretty much every user ever.No* user has ever said to themselves "Ah! The password field! I want to enter the octets 0xC4 0xD6 0xDC! Now, hm, I'm using <checks quickly>... ah, yes, I'm using ISO 8859-1. So, to enter those octets... <looks up ISO 8859-1 table>... ah, I'll need to enter the characters Ä, Ö and Ü to do that.".
Seriously, nobody* has ever done this. What users say to themselves is, "Ah! The password field! My password is Äpfel. I'll type in Äpfel!". And they expect this to work. And they expect it to work whether they're typing in their password on Windows, using Unicode, or some *nix using 8859-1, or whatever. If they're using a system which only supports ASCII (or EBCDIC), they're not going to be able to enter Äpfel in the first place.
Users don't know, don't care, shouldn't have to care and probably never will care about what octets those characters are. They care about words and the characters that make up those words.
* with the possible exception of yourself, since you're campaigning so vehemently for this. If you do do this, then I'm sorry, but I'm pretty sure you're a lone voice, crying in the wilderness. Presumably, crying in octets.
-- Jon Bright Silicon Circus Ltd. http://www.siliconcircus.com