The interaction of spatially small dipoles with nucleons, nuclei is calcula
ted in the DGLAP approximation at the top of HERA energies and found to be
close to the S-channel unitarity limit in the case of the color octet dipol
es. The DGLAP analyses of the current diffractive data appear to support th
is conclusion as they indicate a similar to 30 - 40% probability of the glu
on induced diffraction for Q(2) similar to 4GeV(2). The need for the high-p
recision measurements of the t-dependence of inclusive and exclusive diffra
ction for pinpointing higher twist effects in the gluon sector is emphasize
d. The eA collisions at HERA would provide a strong amplification of the gl
uon densities allowing to reach deep into the regime of nonlinear QCD evolu
tion. Connection between the leading twist nuclear shadowing and leading tw
ist diffraction in ep scattering is explained. The presented model independ
ent results for the nuclear shadowing for light nuclei indicate much larger
shadowing for the gluon sector than for the sea quark sector. It is argued
that HERA in eA mode would be able to discover a number of new phenomena i
ncluding large gluon shadowing, large nonlinearities in parton evolution, s
mall a:color transparency in the vector meson production followed by color
opacity at x less than or equal to .005, large probability of inclusive dif
fraction. Implications for the nucleus-nucleus collisions at LHC are discus
sed as well.