[DNA] draft-narayanan-dna-routers-bcp-00.txt

[Brett Pentland <brett.pentland@eng.monash.edu.au>]

Hi guys,

I have just been reading over the routers BCP document and had a few
comments.

In general, there seems to be some good information in there for
configuring routers with respect to DNA.  I have a vague recollection
of this being mentioned before, but the Summary of Recommendations
section seems to be one of the more important contributions, and
perhaps deserves to be in its own section of the body of the document
rather than relegated to an appendix.

More specific observations/questions:

Section 2.1, para 6:
"Where multiple unicast transmissions for the same destination await
transmission,"... In what situation would this occur?

Section 2.2:
Maybe some of the details of calculation of expected response times
could be moved to an appendix.

Section 2.4:
The two tables only differ in one row.  Perhaps they could be combined
into one.  This would also make comparison easier as the changed values
would be side by side.

Section 2.4, last para:
I don't think the mean delay before receiving an unsolicited RA is 2.25
seconds.  That is the average interarrival time for unsolicited RAs,
isn't it?  If the host always arrived just after an unsolicited RA, then
the mean delay would be 2.25 seconds, but that won't always be the
case and so the mean delay is less.  I think it's just 1.125 seconds,
but I admit I'm not certain.

Section 3.3, para 2:
How was the number three arrived at as a recommendation for the maximum
number of routers on a link?

Section 3.4:
Just personal taste, so ignore if you wish but:
s/avoid the delay by not waiting for the prescribed random time/choose 
not to delay/
and again in the appendix.

Appendix A, para 16:
I think that this should say that a router MAY split the
options across multiple RAs, but SHOULD include one PIO that's common
to all of them.

Cheers,
Brett.