[e2e] Simple Question on TCP Window Size

[Vernon Schryver]

> To: Vernon Schryver <vjs at calcite.rhyolite.com>, end2end-interest at postel.org
> From: "David P. Reed" <dpreed at reed.com>

> >Perhaps you could sneak the ECN bits that arrive with the UDP datagram
> >into the mbuf that holds the source address, then have a modified recvmsg()
> >marked by a new flags bit that would set msg_flags with the ECN bits. ...

> >that still don't even understand setsockopt(, IPPROTO_IP, IP_TOS,), which

> In general it would be more correctly in the spirit of IP's interface if 
> all "sockets" implementations let UDP set all IP options and see all the IP 
> header info on incoming UDP datagrams.  Christian Huitema should have some 
> input on that for at least one NW rainforest company.

Setting IP options on UDP real sockets works, but how do you get all of
the IP header?  Real sockets have getsockopt(0,,IP_OPTIONS...) such as
used in rlogind.c to check for and turn off source routing, but how do
you get other IP header bits?

OH!  Do you mean to use observe that the ECN bits are borrowed
from the IP TOS bits and point out that get/setsockopt(,IPPROTO_IP,IP_TOS)
should just do the right thing already on real sockets?
That's a cool and (now that you mention it) obvious idea.


>                                         Christian Huitema should have some 
> input on that for at least one NW rainforest company.

Given the fact that the BSD code behind Winsock 2.0 had
get/setsockopt(,IPPROTO_IP,IP_TOS) for eons and if I'm right that
Winsock 1.0 had it, that may be a steep hill to climb even for the
redoubtable Christian Huitema.


Vernon Schryver    vjs at rhyolite.com