I have recently been thinking about traversing procedural graphs (e.g. L-systems and possible game state graphs), and have needed to develop ways of traversing them beyond doing stuff to everything (due to cycles and potentially being infinite). This may have been covered already, if so I would be interested in references. I thought the shape of the methods I developed might be of use to the generic community. The types of thing the idea might be useful for is something like giving karma to all of a certain persons friends and also their friends, but only once, in a social networking site. Basically it involves transforming the data type to another that has the information you need and then automatically deconstructing the data type after you have traversed the structure. What I think makes it more interesting is I decided that the functions to construct and deconstruct should be a data structure, and you can have functions to compose them so you could have a traversal controller to avoid loops and one to only go down two steps and add them together so the construction and deconstruction are done in the right order*, so you can combine them together. Some basic examples of the way I am going to code stuff I use for my procedural graphs can be found on my blog http://vrrm.wordpress.com/2009/03/03/unbounded-graphs2/. If I have some spare time, I'll have a look at a way to do this for Generics. Will Pearson * The loop avoidance would add make the data a tuple with a bool and didn't do anything if the Bool had been set to true. The depth controller would make the data a tuple with an Int that got incremented as you go down recursion levels which you then checked to see if it was above 2, in which case you could halt recursion.