UDP Echo and DiffServ bits

[Glen Turner <[e-mail]>]

There appears to be a divergence in vendor practice as to
if UDP Echo (well-known UDP port 7) should set the DiffServ
bits in the repsonse to the same as those in the request.

RFC862 and RFC1349 don't provide much guidance: the wording
of RFC862 implying that only the data needs to be echoed, so
the contents of the IP header are undefined; and RFC1349
stating that the ToS bits may differ between the two
directions of a flow.

I would like to argue that the UDP Echo service should set
the DiffServ bits in the response to those in the request.

This appears to be the most useful setting.  If the DiffServ
code point isn't supported in the reverse direction, then
the DSCP will be policed to 0 and the packet forwarded.  If
the differentiated service is supported, then the echoing of
the DSCP will allow the most probable network path to be tested.

Does anyone have a definitive statement on UDP Echo and DiffServ
in an RFC or ID?

Are there factors that would suggest that the UDP Echo service
should not set the DiffServ bits?

Ta,
Glen

 [PS: I'm aware of the denial of service possibilities with UDP Echo,
      especially one that sets the DiffServ bits to obtain superior
      network performance.  To avoid argument on that topic, let's
      assume that the source IP address of the request is checked
      against an ACL and that the reverse path forwarding check is
      configured.]

-- 
 Glen Turner                                 Network Engineer
 [phone]      Australian Academic and Research Network
 [e-mail]          http://www.aarnet.edu.au/