Steven raised some good points about the dimensionless specializations of quantities. The crux of the problem is that sometimes we would like to be able to specify a formally dimensionless quantity as a heterogeneous quantity : e.g. 1.0*CGS::dyne/SI::newton. If we automatically convert, we lose the ability to represent quantities like this and all dimensionless quantities become equivalent. However, in some cases you would like to preserve the information on the unit system; for trig functions, as an example, you would like to have closure so that acos(cos(quantity<some_angular_unit,Y>)) returns the argument passed to cos. If system information is stripped, this becomes impossible without explicitly specifying the return type of acos. Furthermore, now we can't represent dimensionless quantities for which conversion factors haven't been defined. Now, if we preserve this capability, the system resulting from a dimensionless unit such as 1.5*CGS::dyne/SI::newton becomes ambiguous - should it be an SI dimensionless quantity or a CGS dimensionless quantity. So, even though it is a bit ugly, I think Steven is right that users will just need to explicitly cast heterogeneous dimensionless quantities to homogeneous ones which are well-defined and convertible to the underlying value_type... 1.0*CGS::dyne/SI::newton -> quantity< unit<dimensionless_type,[mixed CGS/SI system]> >, not convertible to double quantity<SI::dimensionless>(1.0*CGS::dyne/SI::newton) -> convertible to double At least this is consistent with the treatment of heterogeneous units (which will be better documented any time now...) throughout the remainder of the library... Matthias