question regarding math.d and magic numbers

[dennis luehring]

i see many many const values for bitmasking in math.d - wouldn't it be 
better to reduce this magic values down to float traits (and start using 
them)?