> production will allow and require such things as "address me formally,
> in the Northeastern dialect of American English, using the diction and
> pronunciation of an upper middle class white female with a college
I strongly disagree with this point. What may be required one day is that
the _administrator_ of a computer system should be able to set it up so
that it'll use certain diction and pronounciation in certain circumstances.
But I don't think that end users will ever want to bother about these
subtleties on any other computer except perhaps their own PC.
> I'm not suggesting creating the whole protocol right now. I am
> suggesting that the protocol should be designed for easy extension to
> handle the (already well known and well understood) subtleties.
I don't think we even need a protocol for specifying language subtleties
right now. What we need is a protocol for specifying language preferences
and we have one already (namely, the Prefer-Language: header and a
'language' command which will be essentially equivalent to it but which
will e.g. allow 'german' and 'deutsch' as aliases for 'de'.)
What IMHO matters now is to
1) Design a good interface between the localize_message function (I mean
the function which turns a message_tag and values for any variables
in the translation string into a human-readable message) and the rest
of Majordomo. This should be designed to be fast, but also
sufficiently flexible to meet any future needs.
2) Hack up a very simple implementation of that localize_message
function which meets the needs of those languages and language
variations for which we currently have volunteers, namely:
German (two versions, namely 'formal' and 'informal'), Spanish,
Portugese and Welsh. This does not need to account for every
subtlety which might become relevant sometime in the future,
because if we have a good interface which meets long-term needs,
it is no problem to adapt the localize_message function to
meet evolving needs.
3) Get a snapshot out which speaks at least two languages.
Please, let us turn the focus of the discussion to the interface mentioned
in point 1) now. I think that if we can reach a consensus on the interface,
probably Hanno will be able to take care of points 2) and 3).