While fixing another problem in the USB code I found
some things which could be cleaned up without too
much risk (imho):
-Explicite specification of polling intervals for interrupt
 descriptors where the descriptor should tell itself
-Allocation of data for transaction translators in the hub code
 where nothing makes use of it
-The former is even wrong: to use multiple transaction translators
 the hub must be configured to an alternative interface.

Is there anyone working on the USB code, or does someone have
an understanding of the issues good enough to review patches?

best regards
Matthias