More intuitive explanation of PMP::Adaptive_sizing_field tolerance parameter?

[darikg]

The documentation says tol is "the error tolerance, used together with curvature to derive target edge length. Lower tolerance values will result in shorter mesh edges."

tol is used in the source code thusly:

const FT vertex_size_sq = 6 * tol / max_absolute_curv - 3 * CGAL::square(tol);

and then vertex_size_sq is compared against the minimum and maximum square edge lengths. Is the tolerance expected to be relative or absolute? It kinda seems like both.