I am unable to persuade route to behave the way I want it to.  Our local
net (which is slightly confused, possibly), looks like this:
	192.129.84.[0-15]	local ethernet
	192.129.84.[16-31]	off-site ethernet network
They're connected by a PPP link from 192.129.84.3 to 192.129.84.17.
So, on .3, we have le0 configured with a netmask of 255.255.255.240.
On .17, le0 is onfigured with a netmask of 255.255.255.240.

.17 immediately realizes that .19 (for instance) is on local ethernet,
and that .4, .8, et al. are across the PPP link.

.3 is incapable of doing this; an example netstat -nr gives
192.129.84.17    192.129.84.3       UH          3    22219      -  ppp0
192.129.84.18    192.129.84.17      UGHS        0        0      -  ppp0
192.129.84.19    192.129.84.17      UHS         0        0      -  le0
192.129.84.22    192.129.84.17      UGHS        0        0      -  le0
192.129.84.23    192.129.84.17      UGHS        0        0      -  le0

Note that .19 is unreachable through le0.  You cannot fix this with route;
at least, I can't.  I don't know why it can get .18 correct, but not .19.

I think route needs the ability to specify, for instance, that ppp0
is to be used for a given route; it's apparently otherwise incapable
of figuring out that using le0 to reach a host connected via ppp will
not work.

We have a (possibly related?) problem where it will periodically get
a gateway of "0.0.75.0".  I don't know where from; I'll get a line like
192.129.84.17    0.0.75.0           UGHS        0        0       - le0
... when .17 is connected through PPP.

Worse, it still gets through ok, although it seems to get some pretty significant
performance hits.

I have disabled routed, because it seemed to make this all a lot worse.

Any advice?

-s