[comp.protocols.appletalk] IPADDRESS vs IPGATEWAY

morgan@JESSICA.STANFORD.EDU (07/22/89)

> Since making this change, atlook now views the K-4 box as an
> IPADDRESS, whereas previously atlook recognized the K-3 box as an
> IPGATEWAY (the "correct" answer).  Is this a bug?  If so, is there a
> solution? More importantly, is this a big deal?  All of our services,
> such as NCSA Telnet and CAP work fine with the box as an IPADDRESS,
> but is this hurting us in some subtle way?

Here's what my experiments show, with KSTAR:

If you sit at a Mac on a LocalTalk and generate NBPLookups (using
CheckNET from Farallon or any other fine product), if you look up
=:=@* (any device of any type in this zone), then Mac/IP|NCSATelnet
users show up as IPADDRESS, other KSTAR-Kboxes in the zone show up as
IPADDRESS (two or three times, one for each of LocalTalk, IPTalk, and
EtherTalk), but one's own KSTAR-Kbox shows up not at all.  Looking up
=:IPGATEWAY@* will produce the appropriate response from the local
KSTAR-Kbox (my.ip.address:IPGATEWAY@*), but none from others in the
zone (this lookup is how Mac/IP and NCSATelnet find the gateway when
they first start up).

If you sit at an Ethernet-attached Mac and generate lookups, or use
atlook from CAP, you see the IPADDRESS responses just the same to a
=:=@* lookup, but no response at all to a =:IPGATEWAY@* lookup.

With KIP (as I recall, we've converted all the local boxes to KSTAR
here), Kboxes will respond to =:=:@ lookups as IPGATEWAY no matter
what, but will not respond to =:IPGATEWAY@* lookups when they come in
via Ethernet.

I think these are just two different strategies to dealing with the
problem that a LocalTalk-attached Mac wants to find *only* its local
IPGATEWAY when it does a =:IPGATEWAY@* lookup.  Both of them slightly
violate AppleTalk good rules of conduct, but perhaps KSTAR's behavior
is a little more confusing.

KSTAR has a configuration option (option 7) that isn't documented in
anything I have, that will cause the Kbox to respond to IPGATEWAY
lookups coming in from the Ethernet.  This could be set if you wanted
to get dynamic IP assignment for an EtherTalked Mac (though this can
be done other ways with MacTCP), or if you wanted to have one Kbox be
the IP gateway for multiple LocalTalks/Ethernets (as somebody was
asking a few messages ago).  Beware, though, that if SU-Mac/IP (and
probably NCSATelnet, too) hears from multiple IPGATEWAYs in response
to a lookup (ie if they're in the same zone), it refuses to operate
since it doesn't know which one to talk to.

 - RL "Bob" Morgan
   Networking Systems
   Stanford

minshall@kinetics.UUCP (Greg Minshall) (07/27/89)

From article <928@wucs1.wustl.edu>, by kahn@wucs1.wustl.edu (Michael Kahn):
> We've recently switched from a K-3 box running KIP to a K-4 box
> running KSTAR
...
> Since making this change, atlook now views the K-4 box as an
> IPADDRESS, whereas previously atlook recognized the K-3 box as an
> IPGATEWAY (the "correct" answer).
> 
> Is this a bug?

Yes, it is a bug.  It is also a bug that KIP only answered IPADDRESS.
Both "should" answer everything they are - WHEN APPROPRIATE.  Bob Morgan,
in another follow-up to this message, indicated that neither KIP nor
K-STAR should reply in certain circumstances which have to do with
where the person doing the NBP lookup is sitting on the network (by the
way Bob, =:=@* *should* return an xxx:IPADDRESS@* answer from localtalk -
but that may be a second bug, sigh).

The reason that two K-boxes replying IPGATEWAY makes the various MacIP
programs die is that each K-box *also* replies to yyy:IPADDRESS@* queries
where "yyy" is NOT one of the IP addresses assigned to the individual K-box.
(This makes some MacIP programs work).

Hope this is useful.

Greg Minshall			Kinetics/Excelan/Novell
minshall@kinetics.com		1-415-947-0998