Does it filter the resulting kerning by checking if the resulting pairs are present or not in the languages covered by the chosen codepage or encoding?

No.
And it does not check to see whether the characters are used in the same language.
Consider Latin 1, which contains, for instance, ccedilla and lslash, which are used exclusively in French and Polish, respectively.
If you put a kern between c and l, and include ccedilla in your c kerning class, and lslash in your l kerning class, then your expanded kerning will include an entry for the ccedilla-lslash combination, although this never occurs in any language.

At least, that's my understanding of the situation.

Some people have gone to the trouble of investigating real-language occurrence of character pairs, and using it to limit the number of kern pairs in their fonts. House Industries and Lucas de Groot, IIRC.

Thanks Nick. I guess for the moment I will continue generating (and checking for collisions) hundred of unnecessary pairs :S

Ramiro.

Ramiro,

the filter just limits the pairs to glyphs which are part of the selected encoding or codepage, i.e. pairs in which one of the glyphs is outside of the encoding or codepage that you select, are not created.

A.