The standard theoretical description of polarized DIS is based
on the use of the DGLAP evolution equations. This approach is theoretically
grounded for the hard kinematics of large x and large Q2. It controls the Q2
-evolution but cannot account for the x -evolution. Instead, it mimics the
x -evolution by ad hoc inserting singular factors x^-a in parton distributions
without any theoretical grounds. Resummation of the leading logarithmic
contributions and merging it with DGLAPP allowed us to describe g1 at
arbitrary x and drop the singular factors at the same time. Then, independently 
of values of x, the region of small Q2 is also out of the reach of
DGLAP. We suggest a simple way to extrapolate DGLAP into the small-Q2
region at any x, arriving thereby at description of g1 at arbitrary x and Q2.