`Johannes L. Braams' wrote
>           - The concept of language attributes is introduced. It is
>             intended to be used for giving the user some control over the
>             use of the features a language definition file provides. It's
>             first use is for the Greek language where the user can select
>             to use the  \pi o\lambda\upsilon\tau o\nu\kappa\acute{o}
>             (``Polutoniko'' or multiaccented) Greek way of typesetting
>             texts. These attributes wil possible find wider use in future
>             releases.

Thanks for this new version, Johannes!

Among the language attributes, there should IMHO be "default encodings."
These encodings would be either a pair of < input encoding, output encoding >
or a pair of lists of encoding, which would enable priorities.
For instance, English should have as a default < ASCII, OT1 >.
French should have < latin1, T1 >, Russian could have < koi8-r, X2 >, etc.
These encodings would then just be triggered when one writes
\selectlanguage{french}, etc. I haven't tried the new version of babel,
so maybe this is already included, and in this case I apologize,
but in versions 3.6, when I want to switch between french, english and
russian in the same document, I have to manually switch the encodings.
And this is a bit cumbersome, especially since it could (and should IMHO)
be automated. I think that when possible,
encodings should be hidden from the user.

Denis Roegel