I can certainly see how a server-based version would be a little problematic now, but that's really the way the industry is going - networks are getting better, unlimited data is becoming a staple of mobile plans (with the iPhone in fact I think it's a subscription requirement), and hardware-wise, processor speed has been largely stagnant the last couple of years while manufacturers have been focusing on things like form factor, screen size, and battery life, along with of course the new 3G chipsets. The vast majority of BlackBerry owners use their BlackBerry primarily for e-mail, so you're unfortunately in the minority in terms of wanting to use our software on your BlackBerry without having internet access.
But certainly my views on this could change, if we release a (very-easy-to-port) touchscreen-less Windows Mobile version and it turns out to be hugely popular that would definitely be an argument for considering a native BlackBerry port.
Thanks for the additional comments on handwriting recognition - we've been looking at some sort of multiple-radical lookup (where as long as you recognize *some* component in a character it doesn't actually have to be the main one, and you can key in more than one to reduce the number of choices for oft-used radicals like water/speech/etc) as a way to improve the radical system's accessibility to beginners. Handwriting recognition is definitely key if you're reading paper documents, so I imagine its heaviest users are people dealing with printed translations, textbooks and the like, but as a demo / "hey, look at that!" feature it's still beyond almost anything else we offer. (even more so with our new handwriting-input flashcard mode that I'll hopefully be posting screenshots of next week)