Re: [Gen-art] Gen-ART Telechat Review of draft-ietf-6man-flow-update-07

[Brian E Carpenter <[e-mail]>]

Pete,

On 2011-07-14 10:16, Pete McCann wrote:
> I am the assigned Gen-ART reviewer for this draft. For background on
> Gen-ART, please see the FAQ at
> < http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.
> 
> Please wait for direction from your document shepherd
> or AD before posting a new version of the draft.
> 
> Document: draft-ietf-6man-flow-update-07
> Reviewer: Pete McCann
> Review Date: 2011-07-13
> IETF LC End Date:
> IESG Telechat date: 2011-07-14
> 
> Summary: One clarification needed
> 
> Major issues: None
> 
> Minor issues:
> 
> It isn't clear to me what the document means by "stateful" vs.
> "stateless" assignment
> of flow labels.  Approximating a uniform distribution with a hash
> function or PRNG
> requires some sort of state such as a seed or sequence number which is updated
> after each assignment, yes? 

No. In fact there's an example of a purely stateless algorithm that requires no
seed or sequence number in the Appendix of draft-ietf-6man-3697bis-06; there just
needs to be enough variability in some of the header fields (typically the destination
address and one of the port numbers).

> Or were you referring to the statefulness
> of intermediate
> routers that identify a flow by port number and then remember the flow
> label assignment
> they chose?

Not exactly - the real distinction is between scenarios where there is a signaling
process to set up a flow with a particular label (for example, Intserv/RSVP
is defined to be able to signal a flow label) and scenarios preferred by the WG
where there is no signaling and no pre-established state anywhere.

Would this change in  Section 2 be OK for you?

OLD:
   It is
   not made clear by the rule that there is an implied distinction
   between stateless models (in which case no assumption may be made)
   and stateful models (in which the router has explicit knowledge).

NEW:
   It is
   not made clear by the rule that there is an implied distinction
   between stateless models (in which there is no signaling, so no specific
   assumption about the meaning of the flow label value can be made)
   and stateful models (in there is signaling and the router has explicit
   knowledge about the label).

      Brian