gettext locale environment variable documentation

[Karl Berry]

Hi Bruno and all,

First, in the "End Users" node of the current gettext.texi, it is said that
setting the LANG envvar is all that is necessary to set the locale.
While I guess that is true, it is a bit misleading.  I think it would be
helpful to also state that the LC_ALL and LC_MESSAGES, and sometimes
LANGUAGE, envvars override LANG, pointing to the "gettext grok" node for
details, I guess.  (Based on that high-level description, I got rather
confused messing about with LANG and seeing no effect, because the other
envvars were set for whatever reason.)

Second, there is an additional wrinkle described in gettext(3)
which is not in gettext.texi as far as I can see, about LANGUAGE.
Namely, as I understand it, if LC_ALL (or LC_MESSAGES) != C, and
LANGUAGE is set, then it is used.  Put another way, if LC_ALL *is* C,
LANGUAGE is ignored.  This is not explained in the gettext grok node,
which just says LANGUAGE is always used first.  (This also managed to
confuse me.)

Third, also in the gettext grok node, there is this item:
   3. `LC_xxx', according to selected locale 
I don't understand why the phrase "according to the selected locale" is
there.  I surmised that xxx meant names like "MESSAGES", e.g.,
"LC_MESSAGES".  How is that locale-dependent?  Does "MESSAGES" get
translated (LC_ANZEIGEN, babelfish tells me :)?!  Can you explain,
please?

It wouldn't surprise me if I'm misunderstanding or stating something
wrong.  Corrections would be greatly appreciated.

Thanks,
Karl

P.S. Maybe also worth mentioning that "POSIX" is the same as "C" for a
locale name?  (At least I believe it is.)  I didn't see this in either
gettext.texi or gettext(3).