both of these sound useful.

> I wanted to integrate a couple of small hacks that I did for a previous
> employer (who is happy for this stuff to see further use).
> 
> The first was to add support for FreeBSD's TCP_NOPUSH socket option.
> This is somewhat similar to Linux's CORK stuff in that it will
> make sure that it only sends out full TCP segments.  If you've 
> provided less than a full segment of data, it'll wait indefinitely
> for you to either give it more data, or turn off the NOPUSH option,
> or close the socket.

When I'm in a weird mood, I get the sense that fsync() on the socket
should also cause a PUSH ;-)

> The second was even smaller and easier.  I set up redirected routes
> so that they automatically went into the route timeout queue, and 
> would disappear after ten minutes or so.  We needed this, because,
> for really bizarre reasons, it wasn't going to be possible to run a 
> route daemon on our system.  It seems to me that there are some
> situations in which this might be useful.  Of course, it'd be
> controlled by a sysctl, and have another sysctl to control the
> timeout period.  And it would be off by default.

yes.  anything learned off the network can be wrong; bogus redirects
should eventually expire; the fact that they don't is a bug..

					- Bill