Why is eachindex faster than 1:n?

Why is Base.OneTo(n) faster than 1:n (in indexing), I read that the lower bound is guaranteed (by the compiler) to be 1, but I don’t get How ? and Why isn’t 1:n's lower bound guaranteed to be 1 ? and why don’t the julia developpers make colon(1, n)::Base.OneTo(n)

Base.OneTo(n) is known from its type to start at 1. 1:n cannot from its type be known to start at 1 (it has the same type as 2:n).

It would be bad if 1:n and 2:n had different types.