[JDEV] charsets (was: Protocol extension?)

[Lindsay.Marshall at newcastle.ac.uk]

> How do the client know the language of a message? "Normal" human usage
> patterns would probably be to write most mesasges in their native
> language and some in english. It seems to me that selecting language
> on every other message will likely be too much of a hassle. Of course, it
> would be possible for every user to enter preferred language in your
> profile in the roster. Intelligent clients could guess that yuo write a
> message to a person in his/her preferred language. That would reduce
> the number of occasions of manual language selection, if not eliminate
> it.

The client simply looks at the default locale and that will tell it the
local default language. It could perhaps negotiate with the server to
see if its default is the same so as to avoid sending it all the time.

> ... or we could pipe the text through spell with varying dictionaries
> and select the one that has least errors. ;)

Automatic langugae recognition is hard - there is a website that tries
to do it and it is not very good.

> All this is a client issue of course, and shouldn't bother the
> protocol/DTD itself. Misuse is no reason not to do things right. Still
> it will be an issue. Any suggestions?

It is *not* jut a client issue. The information has to be embedded
into the protocol so needs to be in the DTD somewhere.

L.
-- 
http://catless.ncl.ac.uk/Lindsay