Path: utzoo!attcan!uunet!mcsun!cernvax!chx400!urz.unibas.ch!doelz
From: doelz@urz.unibas.ch
Newsgroups: comp.sys.sgi
Subject: Re: NQS on iris fed from sun
Message-ID: <1990Oct3.154508.979@urz.unibas.ch>
Date: 3 Oct 90 14:45:08 GMT
References: <1990Oct2.152835@igc.ethz.ch>
Organization: University of Basel, Switzerland
Lines: 27

In article <1990Oct2.152835@igc.ethz.ch>, torda@igc.ethz.ch (Andrew Torda) writes:
> Imagine one has sgi's port of NQS running on an iris and Sterling softwares
> original NQS running on a sun.
Applies also for CRAY and CONVEX NQS. 
> 
> Has anyone managed to persuade the iris to accept requests from the sun ?
> NQS on suns, crays and possibly others seems to put great weight on user id's
> machine id's and the nmapmgr (qmapmgr on cray) program to inform the batch
> system of this mapping.
Called qmapmgr on Convex as well. 
> 
> Sgi's port does not seem to provide such a mapping. What does this mean ?
> Thanks for any advice.
> --
> Andrew Torda, ETH, Zurich

Qmapmgr maps the ethernet location of a machine to the nqs socket. 
The resulting entry is kept in a (binary) database and the entries are 
numbered as integers, the so-called machine id. In order to make things easier, 
SGI is using the internet address. I failed so far to hack the connection 
in between the two because the mid (machine id) cannot be a four byte 
number. Anyone else in netland having had success so far ? 
(Or, anyone of SGI ever thought of commenting on the problem of interconnecting
hardware? :-) Sorry, I appreciate that things are easier if you don't need 
to get just another database to maintain but in this case it would be nice.)

- Reinhard