On Wed, 14 Jan 2009, Rick Jones wrote:

quoted text
> >>How likely is it that the "additional small delay" above would be much 
> >>less than waiting for a read return of zero after a shutdown(SHUT_WR) call?
> > 
> > 
> > I'm not sure I understand what you're getting at.  I did consider doing
> > something like what you suggested, but in the end decided it was simpler
> > to deal with a fully ESTABLISHED connection, than worrying about possible
> > races with a socket being (partially or fully) closed.
> 
> Ostensibly, using a shutdown(SHUT_WR) and then a wait for a recv return 
> of zero would take about the same length of time as polling local 
> connection stats to see that there were no ostensibly unacked data - 
> both will take one RTT right?  and shutdown/read has the added property 
> that it will deal with zero windows automagically.

With the shutdown(SHUT_WR)/read() approach, I would have had to set
a timeout on the read, to handle the case where the peer just went
away, whereas currently I just check elapsed time (I strive to make
nuttcp robust in such cases to allow it to be used reliably within
scripts run for example from cron).

Also, I was (perhaps unncessarily) worried that after the zero read(),
the socket would effectively be closed, and I wasn't sure then about
the reliability of using tcp_info to get the tcpi_total_retrans at that
point.

As with most things, there's more than one way to skin a cat.

						-Bill
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html