hunk ./docs/NOTES 1
+* Algebra.Module
+
+I think it should be a type family rather than a multi-parameter type class.
+My main motivation for multi-paramter type class
+was to allow complex numbers to be a vector space over both real and complex numbers.
+This does not worked well and even more type inference often fails.
+We should just have two different types of complex numbers:
+One complex number type being a vector space over reals
+and another complex type being a vector space over complex numbers.
+
hunk ./docs/NOTES 1
+* Positional: test suite
+
+* Positional and zero
+
+Represent zero with empty mantissa?
+Or better have NonZero type with non-empty mantissa
+and a full number type with optional zero?
+Or something where we can have negative numbers and zero as option?
+Problem is, that we allow negative digits
+and thus even a Positive number type can represent zero and negative numbers.
+
+We might at least define a NonEmptyMantissa type for interim computations,
+like in 'divide'.
+
+* Positional.Fixed
+
+We could derive the base from digit type, e.g.
+ Int32 -> 1000
+ Int64 -> 1000000
+ newtype Integer -> anything
+