IntInf and word operations

[Stephen Weeks]

Does anyone here understand the discussion of representing IntInfs via two's
complement in the basis library spec?  It makes no sense to me.  I am trying to
understand the spec for Word.fromLargeInt, which requires that the IntInf be
(conceptually) represented in twos complement.  But I don't understand what the
representation of an arbitrary precision integer in twos complement is.  How
many bits do you use?