ACTIVE BRANCHES: 

simpler-speeds		: deal with links of different speeds in the network.
			  this branch generalizes the simple breadth-first
			  traversal to a traversal according to a priority
			  queue instead of a list of nodes to visit. the
			  priority of a node is then the lowest link speed
			  on the path from the root to that node.

			  TODO: debug. it seems to do kindof the same thing
			  in that the links are propagated correctly, but
			  something in the algorithm makes nodes receive
			  pruned subtrees that should not have been pruned for
			  those nodes, so destinations in the network get
			  lost. a better diagnosis is needed, and then a
			  solution worked out, if one is possible. it may very
			  well be that there is something fundamentally wrong
			  with the approach.

DORMANT BRANCHES (as in, not likely to make trunk in the foreseeable future):

forest			: pass a list of trees instead of only one tree.
			  the intent is to support multiple disjoint address
			  spaces with one instance of the daemon.

			  dormant because the initiative that spawned it seems
			  dormant itself. 

ipv6			: IPv6 support. this is a big TODO as of yet. choices
			  to be made are whether or not to distinct trees
			  or mandate that for every IPv4 address there is a
			  corresponding IPv6 address and somehow stick them
			  into one tree. also the packets are going to be
			  bigger either way, which is another problem. 

			  also, if anyone wants a stab at this
			  and chooses the distinct-trees option, it might be
			  better to create a new subbranch of 'forest' and
			  cherrypick the lowlevel IPv6 bits from here.

OBSOLETE BRANCHES (as in deleted because not likely to make trunk ever.
	ofcourse this being an svn repository you can always resurrect them
	if need be, which is why the revision it's been deleted is listed):

new_interlink_scheme	: support for a proposed interlink scheme where an AP
			  and its clients share a subnet, instead of the AP
			  having one point-to-point /30 between it and the
			  client.

			  TODO: deal with the fact that two clients are not
			  directly connected even though they're in the same
			  subnet. probably mangle the tree such that it is
			  exactly the same as it is now, and deal with the 
			  route stuff in the route applier or something.

			  deleted as of rev 5190, because this scheme has
			  never been implemented.

speeds			: succeeded by simpler-speeds, because that's, well, 
			  simpler and does the same thing. this is basically
			  an older idea with the same goal. deleted in r5190.