[conex] Comments on ConEx TCP Modifications

[Nandita Dukkipati <[e-mail]>]

Following are a comments as individual contributor on
http://www.ietf.org/id/draft-ietf-conex-tcp-modifications-01.txt and one
comment as chair.

Nandita

1. What changes, if any, are required to accounting when ConEx marked
packets get lost?

2. Sec 3
"A sender which sends different sized packets with unequally distributed
packet sizes should know about reason to do so and thus may be able to
reconstruct the exact number of headers based on this information."
Sentence doesn't read well; please reword.

3. Sections 3.1.1 and 3.1.2
By acked_bytes I assume you mean the cumulatively acked bytes or the change
in snd.una. What happens when there is loss and snd.una does not advance?
Instead, I suggest that you consider using DeliveredData as defined in
http://datatracker.ietf.org/doc/draft-ietf-tcpm-proportional-rate-reduction/?include_text=1
-
it tracks precisely that amount of data delivered to the receiver for each
ACK even in the loss case.

4. Section 3.1.2
"Unfortunately, in a high congestion situation where all packets are CE
marled..."
marled -> marked

5. Section 3.1.2
On the use of 'M'
You won't require to use any variable 'M' when you track precisely the
bytes delivered per ACK using the DeliveredData that I mentioned above.

6. Sec 3.1.2
"In average the sender will sent..." -> "On average the sender will send..."

7. Advanced Compatibility mode
Not clear to me at all what this text is trying to say.

8. Sec 3.2
"Note that the above heuristics delays the ConEx signal by one segment, and
also..."
You mean delay by 1 RTT?

Comment as chair

Section 3 Accounting Congestion
bytes vs. packets: I would like to see consistency across the two drafts
(abstract mechanism, TCP modifications) on the issue of whether ConEx
marking and accounting should be in bytes or in packets. What are the
pros/cons and corner cases for each? What is the final recommendation that
you make? Which of these drafts should make this recommendation - I would
think it's the TCP modifications draft.

If you haven't already, could you please read up the latest Concepts and
Abstract Mechanism draft and communicate with the authors of that document
and make sure we have consistency on bytes vs. packets across the two I-Ds.