It seems to me that the automatic creation of loopback
routes for local IP addresses does not work anymore
in -current.
My boxes are diskless, so some leftover of the DHCP
startup might influence this.
However, my (diskless) 1.3.x boxes have this route set.
Can somebody comment on this?

My real problem is with a second ethernet card:
I want to have it in the same subnet as the primary
interface, but for use only by a selected set of host
routes. (It's for a dedicated connection to the file
server which will eventually be FDDI.)
OSF/1 deals well with this kind of setups, so I thought
it should be possible with any modern 4.4BSD based OS...

The interface is set up with
zelz26# ifconfig ec0 zelux0 netmask 0xffffffff
which leads to the lines
ec0   1500  <Link>        02:60:8c:6b:59:b3    22940     0        1     0     0
ec0   1500  zelux0.zel.kf zelux0               22940     0        1     0     0
in "netstat -i" and
zelux0.zel.kfa-jue link#2             UC          0        0      -  ec0
in the "netstat -r" output. If I issue a
zelz26# route add -host zelas3 -iface zelux0
now, the line
zelas3             zelux0             UHS         0        0      -  ec0
appears in "netstat -r", but the log message
"arp_rtrequest: bad gateway value"
shows up. Any attempt to use the route (eg "ping zelas3")
only leads to
"arpresolve: can't allocate llinfo"

(zelz26 and zelux0 are my local addresses, zelas3 is the
other side of the connection; all machines are in the same
subnet.)

Do I have a chance to get this working?

best regards
Matthias