Pro: users who are accustomed to smart completion try to use it in all
contexts, and they will be annoyed by seeing no variants in languages,
where it isn't implemented.
Contra: In general it's just another separate completion, why should it
work as basic? Why not as class name? Why shouldn't class name
completion work as any of the two remaining? And advanced users will
know exactly which kind of completion they need in particular place, and
won't mix them up.
Offtopic: Class name completion in Diana completes not only class names,
but, for example, fully-qualified XML tag names. Shouldn't it be called
in some other way then? How?