On Fri, Jan 10, 2014 at 3:47 PM, Joel M. Halpern <jmh@xxxxxxxxxxxxxxx> wrote: > The problem with taht paragraph is that it assumes that the UDP encapsulator > knows what the payload is. > It seems to me that one could easily be applying the UDP in any one of a > number of cases where the payload will not be known. > So the requirement in the paragraph seems at best difficult to meet. > > And has been noted, this seems to place an expectation on UDP encapsulated > MPLS that is not present for MPLS itself, when the traffic is not known to > be IP. OK good point - so we invoke the end-to-end argument on MPLS's behalf.