Total charm-pair cross sections in $pp$ collisions are interesting because
they can be calculated to NNLO in QCD without any reference to fragmentation
effects. On the other hand, the fiducial differential charm cross sections
from which the total cross sections must be extrapolated are currently
known to NLO+NLL at most (e.g. FONLL), and must be treated for known effects of
nonuniversal charm fragmentation. A new procedure using the FONLL framework
as input for an empirical parametrization of the data in both shape and
normalization, with all its parameters actually fitted to data, is used to
derive so-called data-driven FONLL (ddFONLL)
parametrizations which can be used to extrapolate the differential
cross sections to total cross sections with minimal bias.
This includes an empirical
treatment of all known non-universal charm fragmentation effects,
in particular for the baryon-to-meson ratio as a function of
transverse momentum.
The total charm-pair cross sections obtained in this way, which supersede
all previous ones obtained using the assumption of charm-fragmentation
universality, are consistent with NNLO predictions, and allow first studies
of their sensitivity e.g. to the charm-quark mass and/or the NNLO gluon PDF at
very low transverse momentum fractions $x$.