Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!zaphod.mps.ohio-state.edu!wuarchive!mit-eddie!bloom-beacon!eru!hagbard!sunic!mcsun!ukc!pyrltd!root44!praxis!tkr
From: tkr@praxis.co.uk (Tim Rylance)
Newsgroups: comp.protocols.appletalk
Subject: Re: Need help building Rutgers CAP on SunOS 4.1
Message-ID: <5399@newton.praxis.co.uk>
Date: 3 Oct 90 20:27:18 GMT
References: <1644@cnca.cnca-cam.fr>
Organization: Praxis, Bath, UK
Lines: 65

scarpell@cnca.cnca-cam.fr (Claude Scarpelli) writes:

>I'm in trouble building Rutgers CAP on SunOS 4.1.

I had the same problems:

  1. aarpd starts OK and acquires a node number
  2. atis starts OK but doesn't discover EtherTalk network number
     (even though it sees RTMP packets from Kboxes)
  3. aarptest always says "nis 0.0"
  4. atistest fails to register a service, as does aufs.

(I have a small network - a single ethernet backbone with 3 Kboxes
running K-Star 7.0, a bunch of Suns, and about 15 Macs with ethernet
cards.  I don't run atalkad.)

I fixed things with the following patches:

*** lib/cap/aarpd/aarpd.c-	Sat Jul 28 22:30:52 1990
--- lib/cap/aarpd/aarpd.c	Tue Oct  2 17:20:56 1990
***************
*** 79,85 ****
    init_fdlistening();  /* low level scheduler */
    etalk_init(&id, FALSE);
  
!   this_node = etalk_mynode(&id);
  
    printf("Acquired node number %d\n", this_node);
  
--- 79,85 ----
    init_fdlistening();  /* low level scheduler */
    etalk_init(&id, FALSE);
  
!   this_node = nis_node = etalk_mynode(&id);
  
    printf("Acquired node number %d\n", this_node);
  
*** etc/atis.c-	Sat Jul 28 08:50:51 1990
--- etc/atis.c	Tue Oct  2 17:24:32 1990
***************
*** 653,658 ****
--- 653,660 ----
    net = *(u_short *)pkt;
    log(7, "Got RTMP pkt net %d from %d.%d", net, addr->net, addr->node);
  
+   if (addr->net == 0) addr->net = net;	/* tkr: must be our net */
+ 
    if (net == addr->net)
      SetMyAddr(addr);
  
Initializing nis_node in aarpd is a "good guess", it would be better to do it
in atis but I couldn't stomach #include "../lib/cap/enacps/sharedpage.h"
(or rearranging the source tree).  This should make the package work on
an isolated EtherTalk network (no routers), though I have not tested this
configuration yet.

The change to atis makes it acquire the network number from the "Sender's
network number" of the first RTMP packet it sees.  "Inside AppleTalk" (1st ed)
says on page 5-12 "this field allows the receiver of the packet to determine
the network number through which the packet was received".

Have I missed something?  Comments and flames welcome...

Tim Rylance
tkr@praxis.co.uk