Reply to comment:

Cecil Ward
I would appreciate some clarification and expansion. Please be kind and forgive my ignorance. Who what where in the stack is choosing values for the flow label? Applications are not necessarily being written or rewritten to be flow-label aware. If they are dual-stacked how can they manage when this feature is not available in IPv4. Transports could use the flow label instead of apps but how would they cope with lack of support in the entity at the remote end or even worse a different and unknown usage scheme for the flow label by the other end? The lack of clarity in the original spec and the lack of a usage scheme self-identifier are problems that imho can't be fixed without the availability of the Tardis. In closed controlled environments these problems are not a factor but kit and software is, I would assume, not being built that only works in such niche environments. Or is it? An RFC that I read recently seemed to present a fairly hopeless picture. Thanks for a useful article. Cecil Ward.