Re: UTF8

[Simon Josefsson <jas%extundo.com@localhost>]

der Mouse <mouse%Rodents.Montreal.QC.CA@localhost> writes:

> (b) That aside, why shouldn't one octet-string system talking to
> another octet-string system be unable to use non-ASCII octets?  If a
> human tells them to do this, it's up to the human to ensure they use
> sufficiently compatible character sets when character sets matter.  But
> such cases are relatively common - if both systems are
> encoding-agnostic, but the user has chosen to use the same character
> set on both, for example.

That paragraph allowed me to understand your position.  I think we
fundamentally disagree.  Here's my view:

The point of requiring that you (and everyone else) use UTF-8 in the
protocol is to make sure humans do not need to go through the hassles
of ensuring they use compatible charsets.  Few people understand these
matters to make an informed choice.  Further, this is an error prone
topic, so it is not reliable to rely on humans to perform the task.

I believe most users expect things to work.  In this case, things
can't work unless the entire world is using the same charset (not
going to happen), or the protocols use charset-labeled fields and
different systems convert as appropriate.  Or, as you would prefer,
users handle charset conversion by themselves.

I'm not sure you realize the importance of the problem -- non-ASCII
username and passwords is not a "nice thing".  It is not something
that you will rarely see in practice, so that educating those special
users about how to convert between charsets themselves is an option.
For the majority of potential users, it is critical that they can use
symbols they are familiar with.  Only a small part of the world uses
languages that are written in the ASCII alphabet.

You might want to read RFC 2277, it discuss these matters more
eloquently than I can hope to do.

Hope this helps,
Simon