Re: [I18nrp] charter -01

[Nico Williams <[e-mail]>]

On Wed, Aug 15, 2018 at 02:42:34PM +0000, Hollenbeck, Scott wrote:
> > Internationalization Directorate (i18ndir) Charter
> 
> "Internationalization" sometimes drifts into areas that I would
> describe as "localization". That is, use in some local-only context.
> Does anyone else see this distinction, and if so, is it something to
> be concerned about?

At Sun we used to speak of I18N, L10N (localization), and G11N
(globalization).

I18N = everything to do with Unicode and other charsets/codesets,
       including whatever on-the-wire negotiation thereof

L10N = everything to do with localization, including how to express
       language choices, locales, and so on, but also including
       on-the-wire language/locale negotiations

G11N = I18N + L10N

For the IETF most of L10N is of limited relevance, in that Internet
protocols that carry localized/localizable strings need to support
language negotiation, but locales are almost right out because we mostly
want to only use Unicode (UTF-8) on the wire.  Except that locales are
in scope for protocols like SSHv2, where data in arbitrary codesets will
be exchanged (even though messages for users from the server are always
in UTF-8) because the protocol carries arbitrary data in "channels".

Because of this we could get away with using I18N to mean G11N.

But still, the fact is that SSHv2's language negotiation is insufficient
for what users need.  Users work around this by sending POSIX LANG/LC_*
environment variable settings and hope the server has matching locales.
This state of affairs is due to the IETF not having a notion of locale,
which in part is also due to shortcomings / non-universality of POSIX
locales (how would one express a locale interoperably between Windows
and POSIX-ish systems?).  In principle this can cause significant pain,
but in actuality it's mostly a non-problem -- a theoretical problem that
mostly only bothers _me_ :)  We're probably not going to remediate this
because SSHv2 is pretty much the only Internet protocol that needs this,
and it is done (SECSH WG has concluded), and no one is complaining
enough to bother fixing it.

Nico
--