T9 support isnt' extremely hard, its nearly impossible considering T9 is a proprietary implementation. There has, however, already been work done in the past on predictive text that works similarly to T9, but it's rather slow, which is why I imagine it hasn't been merged into trunk.

I also noticed that the implementation didn't technically come up with all possible combinations, though I suspect this is on purpose, considering my implementation whcih does do so is far slower.

@ossipena: sorry for the lack of contributions, I have been so busy with my other projects that I kind of left vertsms alone (though you seem to be managing just fine without my help ;-)).