If I use vecS for EdgeList and/or VertexList, whenever I remove a edge/vertex this also affects any vertex/edge with a higher numbered descriptor, i.e. removing vertex 0 results in subtracting 1 from every other vertex descriptor. I'm making a graphical editor for a structure that is represented as a graph, and to quickly determine which vertex/edge is selected and should be edited in any way I figure the best way is to keep a copy of the edge/vertex descriptor with the graphical object that is supposed to represent just that edge/vertex. If I use vecS this soon becomes a real pain if I have to update edge/vertex descriptors everywhere whenever I remove some part of the graph. I could of course add my own immutable descriptors as properties in the graph, but that just reverses the problem - I'll have to map these back to the 'real' descriptors when I want to modify/access the graph. I assume I can use listS to store both edges and vertices to solve this problem in that the descriptors become pointers which are not disturbed when I remove a vertex/edge? What about setS? If I go with listS I end up with another concern regarding the performance benchmarks on traversal in the BGL book. Is the code used to create those benchmarks available anywhere so that I can do a re-evaluation with later compiler versions? I hope someone can shed some light here. Thanks, -- Tarjei