Path: utzoo!utgpu!watserv1!watmath!att!pacbell.com!mips!news.cs.indiana.edu!ariel.unm.edu!ariel.unm.edu!dd
From: dd@ariel.unm.edu (Don Doerner)
Newsgroups: comp.protocols.appletalk
Subject: Re: MacTCP on LocalTalk'ed machine
Message-ID: <1991Jan5.165854.9675@ariel.unm.edu>
Date: 5 Jan 91 16:58:54 GMT
Sender: news@ariel.unm.edu (News supported software)
Organization: University of New Mexico
Lines: 77
References:<1991Jan3.000106.19583@ncsuvx.ncsu.edu> <278368CB.4E9F@intercon.com> <1991Jan3.225522.21430@ncsuvx.ncsu.edu> <17219@brahms.udel.edu>

In article <17219@brahms.udel.edu> garrett@brahms.udel.edu (Joel Garrett) 
writes:

>>the FastPath.  Thus, this seems the most
>>likely culprit. He's running gateway
>>code KSTAR that came with his FastPath 4.
>>I'm running the code that came with
>>my PROM upgrade (3.0), is this KSTAR?
>>Any ideas where to get KSTAR code that
>>is still compatible with a FP-3?  Thanks.

> There was someone out on the net that said
> something to the effect that there 
> was a way to patch MacTCP to work with
> non-KIP-style routers (ie KFPS-2 & 3s)
> I sent a note to the guy, but never
> heard anything back from him.  According to
> Shiva (the new keepers of FastPath
> technology) the only solution available is
> to upgrade your boxes to KFPS-4s for $999
> or run KIP if you have a Sun or other
> atalkad-capable machines (does anyone
> know of a list of atalkad-capable systems?
> We don't have any Suns on our local network
> (but we do have a few Iris 4Ds and
> a few other equally un-BSDish machines
> such as HP9000s and old Iris 3130s and
> we've never had any luck bringing up
> atalkad on any of these hosts without
> what looked to be more work than we were
> willing to handle...)

> Can anyone else shed some more light on this matter?

Maybe.  The PROM code in all FastPaths is really silent (also robust, also 
stupid) router code, with a minimum of functionality.  In particular, the 
PROM code in all FastPaths will support the static addressing mode of 
MacTCP, or any other piece of TCP/IP code (e.g. NCSA Telnet), *but* *will* 
*not* support the server mode of MacTCP (a.k.a. the "dynamic" mode of NCSA 
telnet), wherein the driver solicits an available IP address from the 
Kinetics box.  The PROM code in the Kinetics box just simply does not have 
the smarts to perform this function.

So, if you *must* run the PROM code for some reason, you *must* use static 
addressing.

Kinetics' (now Shiva's) KSTAR code was originally based upon some code 
called 'KIP' which was developed, and is still available, at Berkeley (I 
think you can FTP anon to berkeley.edu and get to the right machine - hunt 
around, and if you can't find it mail me, I will try to conjure up a 
better location hint).  KSTAR does lots of things that KIP does not do - 
DECNET encapsulation, ... ad nauseum.   However, if you can't run the 
KSTAR code, I believe you can run the KIP code on FastPaths 2 and 3 (but I 
am not going to bet anything I really care about on that).  And I believe 
it *does* support MacTCP server style address allocation.

BTW, MacTCP dynamic address allocation is an algorithm for assigning an IP 
address on the basis of the AppleTalk address for the node, and is 
basically useless except in an isolated collection of Ethernetted Macs, as 
far as I can figure...

Also, I believe that the single largest bug in the MacTCP 1.0.0 code was a 
complete failure to support SCSI Ethernet interfaces (MacTCP 1.0.1 is only 
marginally better).  Point being that I am not sure that it is imperative 
for you to upgrade your MacTCP unless you anticipate using it w/ SCSI 
E'net interfaces...

Anyway, that's about all I know about this situation - but I will be happy 
to help answer e-mail queries.  We are running KFPS4s+KSTAR, PhoneNet, 
Ethernet, MacTCP, and a small army of applications (including the 
hypercard stack that I am using to respond to this posting!) successfully.
I can look up any facet of what we are doing, and let you know...



Don Doerner, Communication Manager
University of New Mexico CIRT