TCP-group 1992
[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
ARP publish problem.
- To: tcp-group@ucsd.edu
- Subject: ARP publish problem.
- From: jmd@cube.handheld.com (Jim De Arras)
- Date: Tue, 7 Jan 92 10:14:02 EST
Here's the situation: I have a class C network, and one thinnet
cable. I run TCP/IP, Internet linked through a proteon router. I
also run NetWare LITE on the came cable. And Appletalk (Ethertalk).
All this runs fine, and is just background material.
I have one machine that runs DV, a ham BBS, and NOS, with a 44.
address. Due to lack of slots, and ints, that machine had only one
3c503 card in it, which runs Netware lite, loaded before DV. That
machine used to reach the Internet via a SLIP connection to another
PC, 'Emperor', which had two ethernet cards, one to support Netware
Lite, and one for TCP/IP. Well, DV's latest supports IPX clients in
a window, so I added ipxpkt to the NOS window on the BBS(and hit the
3 driver limit, darn!), to NOS in Emperor. After correcting the
routes table, all was restored to it's former glory, except MUCH
faster.
Now here's the problem. I want to give the HAM NOS IPXPKT port a
class 'C' address from my local domain. I don't want to subnet for
one (or possibly 2) machines that will use IPXPKT. If I do that, and
change the route table in emperor to the new address, emperor to ham
nos works fine. The RIP broadcasts over the local coax convince some
of the other machines to route through emperor for that new address.
However, other machines ignore the odd routeing for an address that
should be on the coax 'native', since there's no subnetting. Proxy
ARP sounded like the answer, I used it to pass a few local 'C'
addresses over a Gracillis link (on a commercial freq!) before, and
it worked fine! However, this time, both packet drivers in emperor
are of the 10mb ethernet class, even though the ipxpkt driver is
really lying. So if I put an ARP PUBLISH entry in emperor, he loses
the 'real' arp entry for HAM NOS, and cannot reach it any more.
The simplest solution (other than subnetting, which I resist for one
machine!) seems to me to be to re-assemble ipxpkt, and tell it to
report it's another class of interface, such as arcnet. But my
problem is that I don't understand the packet interface well enough
to know if I can get away with such a simple 'fix'. If the code was
written to fake ethernet, can it be talked to by NOS, if NOS thinks
it's ARCNET? Or should I try to modify both, and create a new class
of driver that's really the same as ethernet, but has a differet
class #? And would just that permit two ARP entries for the same IP
address?
Or is there some quick and dirty solution I'm missing, such as some
encap trick that'll allow emperor to publish the address, and route
it, too?
Any advice welcome!
Jim De Arras
WA4ONG
wa4ong@wa4ong.ampr.org
jmd@handheld.com