Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!husc6!think!ames!haven!decuac!decuac.dec.com!avolio
From: avolio@decuac.dec.com (Frederick M. Avolio)
Newsgroups: comp.unix.ultrix
Subject: Re: Ultrix resolver
Keywords: DEC 3100, Resolver
Message-ID: <2640@decuac.DEC.COM>
Date: 6 Apr 89 14:20:40 GMT
References: <2342@helios.ee.lbl.gov> <124@crltrx.crl.dec.com> <774@fcs280s.ncifcrf.gov>
Sender: news@decuac.DEC.COM
Organization: Digital Equipment Corp., Software Services, Maryland
Lines: 45


In article <2342@helios.ee.lbl.gov>, envbvs@epb2.lbl.gov (Brian V. Smith) writes:
> What we would like to do is have one of our machines (I'll call it our server)
> running bind to get it's host information from name servers and our other machines run Yellow Pages to query our server for host information in the hope of reducing network traffic and load on the name servers.
> 
> We are currently running YP on all of our machines, using the server as  the YP master.
> 
> The question is, do we just put the following in our server /etc/svorder file:
> 
> bind
> yp
> local
> 
> And have the clients with the /etc/svcorder file:
> 
> yp
> local
> 
> Will the yellow pages programs on our server get their information from bind
> or will they only use the /etc/hosts file?


YP will only us the YP hosts file. Set up the client systems as resolver only
systems.  Have them set up to only query your server system.  Have that
set up as a caching server (or a master server if you are in control of 
your domain).  The others are just slave servers.  They needn't run
named.  Just make sure that have /etc/resolv.conf pointing at your
server.

So... Server should have

bind
yp

Server should run named.

Clients should have the same...

bind
yp

Clients could run named as a slave system.  Clients could also forgo named
and just have /etc/resolv.config point to the server system.

I trust I make myself obscure?