Draft: draft-ietf-sipping-overload-reqs-01
Reviewer: Spencer Dawkins
Review Date: Jan 11, 2007
Review Deadline: Jan 23, 2007
Status: Post-WGLC

Summary: This draft is even MORE basically ready for publication :-)

Two minor points...


> Thanks for the review comments, Spencer. And to you also I owe an 
> apology for the very tardy and delinquint response and document 
> update. Responses below.
>
> Spencer Dawkins wrote:
>> This draft seems to sidestep interaction with TCP (and SCTP) 
>> retransmission behavior. The current focus on UDP is the right place 
>> to start, but perhaps it's worth looking at non-datagram transports 
>> in a separate draft, and/or at some point in the future.
>
> great point. The discussion in section 4.1 is heavily skewed towards UDP. 
> Its an unstated assumption in fact. I've fixed this, so that it 
> explicitly says UDP and indicates how TCP is impacted. Generally, TCP 
> is better, but you can still have retransmits of segments if the 
> kernel can't acknowledge them fast enough.

The draft now points out relevant TCP behaviors (where it implicitly assumed UDP in most of the discussion). That's fine. It might be useful to say someplace that SCTP behaviors relevant to this draft are modeled on TCP - the string "SCTP" doesn't appear in the draft at all.

> I fixed the nits below.
>
> Thanks,
> Jonathan R.

In my original review, I also had a second question:

"Is there any reason to mention interaction with secure transports?"

just below the point you responded to. I didn't see any changes to the draft that responded to this point (which would be fine, if the answer was "no"), or any text in your e-mail that responded to it. It probably got deleted when you snipped the rest of my review in your reply!