I don't know of any literature. My approach has been to load the node data into Python dictionaries and then write code that traverse those dictionaries.
Basically one dictionary is keyed by toNode values (either the TO_NODE value from the attribute table or end point x,y coordinate pairs)and each key then stores the OID(s) and fromNode(s). The other is keyed by fromNode and each key then stores the OID(s) and toNode(s). The technique then is to alternate through the keys in the dictionaries, getting a node, and then looking up the node that are attached (keeping track of the OIDs along the way), rinse and repeat.