Such a branch might also be open for a number of other improvements - like integrating other algorithms that have been in waiting in the wings.
Are there any particular algorithms you have in mind? We've been slowly integrating contributed algorithms into the trunk. The limitation has basically been that we need to find people with time to review the algorithms, tests, documentation, etc. before merging any code.
Hmm.... I just update the trunk and realized there's already a planarity algorithm in the library - and the cycle ratio code. Oops. Maybe this would be a good opportunity to look thru the wishlist on trac and see if there are any other algorithms people are interested in.
No, I doubt it. We've talked offline a few times about building "version 2" of the Boost Graph Library, which would include many cosmetic cleanups, simplified interfaces, and probably remove many of the hacks for older compilers. Such an upgrade would be mostly backward-compatible and would not need a re-review.
I was thinking something a little more large-scale. There are some concepts in my recent work that could be back-ported (mostly to the adjacency list classes) and some new code that makes working with exterior properties a great deal easier (and cleaner). I'm also developing some ideas - I haven't put them to code - on specialized graph concepts (e.g., weighted graphs). Basically, these just simplify reasoning about the graph interface, when you need to provide property maps and when you don't. Andrew Sutton asutton@cs.kent.edu