Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!zaphod.mps.ohio-state.edu!usc!apple!rutgers!mcnc!decvax.dec.com!ima!minya!jc
From: jc@minya.UUCP (John Chambers)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Dealing with systems without nameservice.
Message-ID: <386@minya.UUCP>
Date: 1 Jun 90 02:57:18 GMT
References: <9005230215.AA10732@ucbvax.Berkeley.EDU>
Lines: 51

In article <9005230215.AA10732@ucbvax.Berkeley.EDU>, 08071TCP@MSU.BITNET (Doug Nelson) writes:
> From Stan Stead:
> >        We are running 4.3 BSD.  We currently use nameservice.  However,
> >a site that we wish to exchange mail and files, while connected to the
> >internet, does not use nameservice.  We can access the site by internet
> >number, but not by name.  Is there anyway to "hot-wire" an address
> >(perhaps via /etc/hosts).  Currently, the only way /etc/hosts is read
> >is if named is not running.
> 
> The real answer is for them to set up name service.  I can't think of
> any good reason for sites attached to the internet to not be running
> name service, or to find a site that can run name service for them.

Gee, I can think of several reasons off the top of my head.  For instance,
I'm working in a network testing lab in which we are constantly fiddling
with the wiring, setting up new networks for a few hours, interconnecting
them at random, changing host names and IP addresses, and all the other
things you need to do to test stuff.  It's not even vaguely reasonable
to expect a nameserver administrator to keep track of all our fiddlings,
and usually we only tell them about the couple of "main" machines that
hold the source, or where we want to send/receive mail.  The rest we take
care of by editing /etc/hosts as needed.

An example of how difficult things can be is our couple of DOS machines
with tcp/ip.  They don't have a local hosts file, and must get all addresses
from a nameserver.  The result is that tests on DOS are *much* slower, since
we must go through a nameserver, which means telnetting to the nameserver,
editing files, and all that.

Another nameserver gotcha turned up a couple of months ago, when people
decided to rename and renumber all of the "main" machines in the lab.
The nameserver had the wrong IP address (two digits interchanged) for
the mainest of the machines (the one with all the source on it), and 
it took about a month to persuade the (badly overworked) administrator
of the nameserver to correct the typo.  /etc/hosts came in very handy
during this period.

Yeah, I know, these are all pathological cases that "shouldn't happen"
to many users.  My main response to that sort of criticism is that I've
never yet worked on a machine that wasn't unusual in some major respect.
(I keep looking for one; I assume they must exist somewhere... ;-)  If
you want to get your job done, you often have to find ways around all 
the broken software, and nameservers are often broken with respect to
some subset of the machines on the network.  So you use it if you can,
and learn to live without it when you must.

-- 
John Chambers ...!{harvard,ima,mit-eddie}!minya!jc
--
If you're not part of the solution, you're part of the precipitate.
[Kiel oni ^ci tiun diras esperante?]