[Tutor] round() errors

[Christian Mascher]

Hi All,

I don't understand some response from the round()-function. 

>>> round(1.4783321399,2)
1.48
>>> round(1.95583,2) # this is the ratio of EURO to DM
1.96

so far, so good, but ...

>>> round(10*1.95583,2) 
19.559999999999999
>>> round(20*1.95583,2)
39.119999999999997
>>> a=20*1.95583
>>> a
39.116599999999998
>>> round(a,2)
39.119999999999997

I recall reading somebody explain this behaviour with the problem of
representing base-10 float-numbers in a binary system. Still, I ask
myself, why can't round() be smart enough to handle this on its own
(especially the last case)?  
Since you can do

>>> print '%.2f' %a
39.12
>>> 

why can't round() do the same? Is there any reason?

Christian