[illumos-Developer] DESIGN REVIEW & CODE REVIEW: 917, compliments of Joyent

[Dan McDonald]

Last week I mentioned *four* Joyent-contributed TCP/IP fixes.  Here's #3:

	http://www.kebe.com/~danmcd/webrevs/917/

This was contributed by Jerry Jelinek.  My own review is that the fix itself
is mechanically fine, and I have no review comments on that front.

What I would like in addition to one other reviewer is something not unlike
what Bryan mailed out last week with 915 and 916 discussion:  the analysis
that lead to the change.

I see that the former constant TCP_ISS_INCR is now an ndd(1m)-tunable
variable, with an allowable range from 1 to TCP_ISS_INCR.  I'd like to know:

	- What happened in the field to inspire this?

	- Why a range of 1..TCP_ISS_INCR?

	- Does this fix work with other settings of tcp_strong_iss?
	  We default to 1 (random increment), and have 0 (constant
	  increment) or 2 (MD5 hash + random increment).  Apart from my
	  own feelings that we should default to '2', did folks test these
	  changes on systems with tcp_strong_iss set to different values?

Thanks!
Dan