[Iccrg] feedback on draft-irtf-tmrg-metrics?

[Eric Coe]

Sally Floyd wrote:
> "Metrics for the Evaluation of Congestion Control Mechanisms",
>  internet-draft draft-irtf-tmrg-metrics-07.txt, is available from:
>   http://www.icir.org/tmrg/draft-irtf-tmrg-metrics-07.txt
>   http://www.icir.org/tmrg/draft-irtf-tmrg-metrics-07.ps
> 
> This draft has finished review within the TMRG (Transport
> Modeling Research Group), and is ready to be passed to
> the IRTF for review and publication.  However, before I pass
> this to the IRTF, this is to pass it by ICCRG for feedback.
> If anyone would like to read it and give feedback, feedback
> would be appreciated in the next two weeks (that is,
> by March 6).  After that, I will forward the draft to IRTF
> for review and publication as an Informational RFC.

Comments on the above draft.

o  TYPO: Page 5 under revision "draft-irtf-tmrg-metrics-01a.txt"
      second bullet item should be "Sean Moore" not Seam Moore.

o  Page 14 Paragraph BEFORE section 3.4 first sentence.
    Currently reads:
    "...concern for environments with high-bandwidth flows."

    Should it read:
    "...concern for environments with high-bandwidth delay flows."

    I think it's the tricky combo of high-bandwidth and long delay that
    makes convergence to fairness problematic. Thoughts?

o  Page 14 Paragraph BEFORE section 3.4 second sentence.
    Currently reads:
    "...between two existing flows after a sudden change such as a 
change in link capacity on a wireless link."

    Suggestion to be more explicitly general (really):
    "...between two existing flows after a sudden change in network 
[characteristics|conditions], such as a change in link capacity on a 
wireless link."


o Page 15 Defining Goodput. First sentence
    "...is defined as the fraction of useful data from all of the data
delivered."

    Should this be:
    "...is defined as the fraction of useful data from all of the data 
sent."

     Since the following sentence further defines high goodput as
efficient use of spectrum... data can be sent, and then NOT delivered
(for various reasons) which effects things like efficient spectrum use 
and buffer occupancy. The first definition does not fully capture the 
effect.


o Page 15 Section 3.5  Title and concept of "failures"
     I'm not sure that "failures" makes sense here. This whole section
really deals with "Misbehaving Users". Looking at the examples given:

    o "Failures of the routers in using explicit feedback to end-nodes
       or failures of end-nodes to follow the prescribed protocols."

If the routers are suppose to provide explicit feedback and they are 
not, they are "misbehaving" and if end-nodes are not following 
prescribed protocols then they are also "misbehaving". Failure to 
"follow" protocol standards is a form of misbehaving. I tend to think of 
failures as more in the terms of "dead nodes", maybe more like faults.

If instead we really mean "non-capable nodes" e.g. non-XCP routers on a 
XCP path. Then maybe a title more like "Robustness to Non-capable and to 
  Misbehaving Users"




-eric