Different form of representation of a skeleton. Rather than focusing on the branch intersections, in a curvilinear skeleton, centers on the branches themselves, here we call them channels. The structure basically consists in a set of points and associated radius (of the largest sphere centered at that point and contained in the original structure, from which the skeleton is derived), and a set of channels defined as subsets of these points.
This data structure is adapted for caracterising 'fibrous' materials, i.e. that are comprised of elements that are much longer than they are wide. Things like textiles, hairballs, fibre composites, vascular, trabecular, or neural networks, etc.
More specifically, there are two main substructures: the set of all the points that define the network, their coordinate and associated float value (the radius mentioned earlier), and the set of channels, which are subsets of pointers to the first set. The save format is text-based (in ASCII), so you can open it and have a look.
For viewing, use the Channel_Network_To_Lines module.