Click on our Sponsors to help Support SunWorld
Letters to the Editor

February letters to the editor

February  1997
[Next story]
[Table of Contents]
Subscribe to SunWorld, it's free!

Mail this
article to
a friend

February letters to the editor

E-mail mystery solved?

Last month, Adrian Corbuleanu wrote SunWorld detailing a problem he was having with the "Local Configuration Error." Following are two letters from readers with fixes to this problem.

Fixes to Smarthost

Regarding the letter to the editor on a problem with e-mail (local configuration error - mail loops back to myself).

Most likely, he has defined a Smarthost that is within his systems domain. The only way he'll get mail on the system is if he has it addressed specifically to where "system" maps to the variable "w" and there is a rule that looks for it. Hard to diagnose this on the fly, but he can send me e-mail directly. Some suggestions:
-- get sendmail-8.8.4
-- use sendmail in test mode to see your variables, e.g. /usr/lib/sendmail -bt -d35.9

Carole Fennelly
Wizard's Keys Corp.

Does Sendmail know about this server?

The error "Local configuration error - mail loops back to myself" is normally caused by a machine that is the primary MX (i.e., the main mail machine) for a domain sendmail knows nothing about. If the machine really should be receiving e-mail for the domain, then this can be easily solved with the addition of the domain to class w, the list of address which is considered local by sendmail.

Thus, if the following is in the DNS:     IN MX 5
                    IN MX 10
This means that the machine is the primary MX host for, then the following can be added to


This adds to the list of domains sees as local. This means that mail for and end up in the same mail box.

The letter doesn't really go into enough detail with the problem though. He doesn't say if he is trying to receive mail for another domain, if the local mail is failing, etc. From the mention of a smart host it may be the case that he has configured all e-mail to be redirected to a central mail hub. Thus, the following could be happening:

e-mail -> indis -> mail hub
Thus, the mail hub is getting e-mail for foo@indis and doesn't know how to handle it, creating the error.

More information is really required to tell exactly what is wrong.

It may be worth checking out the sendmail home page at as well. It contains pointers to a number of useful resources, including the sendmail FAQ. There is also the newsgroup comp.mail.sendmail. The O'Reilly Sendmail book is also extremely useful if you're doing anything with Sendmail. Having the bat by your side, if only for moral support, is almost required when hacking For more details on the Bat see

Incidentally, the O'Reilly Sendmail book is called the bat book because of the bat on the front cover. Most of the O'Reilly Nutshell books are now in this fashion.

Simon Burr

This seems to be a FAQ. The error "local configuration error" and "mail loops back to myself" happens when in the file you do not have an entry for Cw. The file should have included: where is the machine which should be receiving mail.

If the is configured as:
Fw-o /etc/

then the file /etc/ should contain the name(s) of the server that will receive the mail.

Duncan Vella

Solving the mystery that's PPP


Also last month, Scot Corrie of Riverside Publishing, a division of Houghton Mifflin, wrote about a problem he was having with PPP. Again, you responded. Here's what one reader had to say to help Scot through his dilemma, plus he has a question of his own.

Tips for PPP on SPARCs

I found the following document to be very useful in setting up the PPP connection between my UltraSparc and my SPARC5:

However, I have a problem with my connection that this document does not address. I use two 33.6 Kbps U.S. Robotics modems for this connection (one on each end), yet my PPP connection is much slower than it should be. I believe the modems are working correctly (i.e., tip is fast). Do any of your readers have any suggestions? I have tried the obvious solutions such as making sure that data compression is enabled on both ends, serial port rates are 38400, etc.

Ken Sembach
Johns Hopkins University

Disabling root logins

To Peter Galvin:

We have a requirement to disable root logins across the network. We are using Solaris 2.5.1/SunOS 5.5.1 with the Common Desk Top environment V1.0.2. In order to limit root access, we modified the /etc/default/login file by commenting out the CONSOLE=/dev/console line and adding CONSOLE= line. Under Solaris 2.4, this would eliminate root access from the chooser screen; however, under the CDE chooser screen, we may still login as root to the desired host. Do you know how to eliminate root access across the network when using CDE?

Frank McNally

Peter Galvin responds:


I don't have a CDE machine to play around with. I've put some feelers out to folks who might know and I'll let you know what I hear back.

If you really want to secure the machine, you could just disable telnet, ftp, and all the r* daemons fired up in inetd.conf. Of course then no one would be able to login remotely. Alternately, you could use the tcp-wrappers (free) suite of control programs. They intervene between inetd and the daemons and imposing logging and access control lists to limit who can access the daemons. Very easy to install and use. The first step is to figure out how CDE accesses the remote machine (what daemon it uses) -- you could determine this by connecting to a remote machine via CDE and determining which daemon it's talking to. It might not even be telnet'd.


How to deny service

To Peter Galvin:

How can I deny a user access if he is already logged in? Graham Bolton

Peter Gavin responds:

Once a user is logged in, the only way to forcibly remove him from the system is to kill the processes associated with him.
-- Peter

Question about XDM

I have a question about XDM and am hopeful that you will be able to shed some light on it. Since XDM does not run a login shell it does not read users .profile which poses a big problem if you want to run XDM and yet source .profile. Is there any way to make XDm read .profile? I have tried to add . ~HOME/.profile in users .xsession as well as in /usr/openwin/lib/xdm/Xsession but to no avail. I would appreciate any help!

Syed A Ali

Readers -- can you assist Syed?

Kudos for e-mail notification

Sending periodic e-mail reminders to me as a subscriber is a wonderful idea. I would never stop to browse the magazine without those reminders -- I always see some interesting tidbit in that e-mail that makes me take the time to check out the article.

Name and affiliation withheld at author's request

More on e-mail notifications

I recently subscribed to SunWorld and find myself reading the next issue before receiving a notification from SunWorld.

How is this possible? At what point do you send out your notifications for new issues/news? Is it in sync with the publication of the new issue or is there a delay for a few days/weeks?

Jacques Rall


Generally, the e-mail notification of the new issue goes out shortly after the issue goes live. While that's the ideal situation, it doesn't always happen. In December, the mid-month update letter went out several hours before the server had the live update, and I ended up with lots of mail from folks telling me about our "bad links." These links became live as soon as the server caught up with the mail. I try to keep that to a minimum.

I hope this helps. Thanks for reading SunWorld.

Stephen Lawton

Find help from Sun

The best place I've found to get help on anything related to a Sun machine is the Sun Managers' List. The address of the Sun Managers list is I forget if you subscribe through majordomo or -request. I know you need to send "add email@address" to subscribe and "delete email@address" to unsubscribe.

In a lot of situations, I've found the answers to be more complete than those from SunService -- they often accurately solve problems SunService couldn't find a solution to.

Laurent Duperval
Systems Analyst
Grafnetix Systems Inc.

In hot water over the C compiler for Solaris

I can't believe all the hype for Solaris -- it's still a memory hog and doesn't come with a C compiler. I went to a lot of trouble to sell my company on more Sun boxes. I set up a Web server to show what we could do. Low and behold, I went to use the C compiler and guess what -- there isn't one. When I called our Sun rep (after five years of using and paying for Sun equipment and support), I was told that the C compiler would cost extra. Congrats. This alone makes Sun unique. At least all the other Unix vendors hide the cost in the price of the operating system. Not good for Sun and worse for me. After pushing Sun I have to eat crow and tell my company Sun is not the choice for future growth.

Kevin Holtz

Navy catches big fish: answer about fatal error

We run the Adrian performance software on many Sun machines at the U. S. Naval Academy. We recently installed Adrian on an Ultra Enterprise 4000 running Solaris 2.5.1. This machine does have a FDDI attached. gives this error: Fatal: member: txunderruns vanished!: Near line 201 Any help would be greatly appreciated.


Adrian Cockroft responds:

Nice to hear that the software is useful.


If you revisit this SE2.5.0.2 download page, you will find a patch for this problem (FDDI5.0 on 2.5.1 doesn't have a metric that FDDI4.0 defined).

Bill's Bookshelf: Andy Grove's Only the Paranoid Survive

To Bill Rosenblatt:

I enjoyed your "Only the Paranoid Survive" column. I would like to apply your very discerning "Inside the Tornado" commentary to it, if you don't mind.

As you might say, the "Internet/Intranet/Java Paradigm" doesn't address the real world situation -- the internet/intranet IS client/server. The phenomenon is the explosion of multimedia interest into the interactive Web by millions of desktop users hungry for the latest and greatest way to connect with their computers.

Confusing the discipline afforded by the Web's need to reach the greatest audience with "cross-platform" Java-mania obscures the reality that the future is components, not languages (if your paradigm is correct, Apple should be concentrating on a Java VM and Web browser, and forget all those OS problems).

Web-authored, page-centric "programming," marrying user intelligence and business sense with "industrialized" componet-enabled technologies will be marketed to millions of knowledge workers in all areas.

The real "tornado" has struck, i.e. the PC industry has hit the enterprise -- barely a year ago. As George Soros says, communications has surplanted computing. And fueled by the big money of business -- what a sight!

The Web is both a by-product and avenue of this phenomenon. Expect the "tornado" to make a few more leaps before its path, and destruction (and creation), can be determined.

John Hoagland

Bill responds:

I agree that the course of the Web industry is not chartable at this time -- and I would add that anyone who claims to know where it's going is full of it :-).

I would argue that applying Geoff Moore's lifecycle metaphor (bowling alley, tornado, etc.) to the emerging Web market is problematic precisely because it's too broad. A "tornado" technology is supposed to be a solution for some fairly specific business process for end-users. The Web has not reached that point yet -- it's more like a set of "atmospheric conditions" that is favorable to the formation of tornadoes. Any tornadoes that come out of the Web world will be much more specific applications of Web technology, such as perhaps microtransaction applications in Java.

Regarding your point that intranet/Internet is client/server, I won't's a matter of whose label set you subscribe to. In a generic sense, of course you're right -- there are clients and servers in the i*net world. But to many people, "client/server" has come to mean something much more specific: the two-tier paradigm with the fat client (Windows, Mac, etc.) and the relational database (Oracle, Informix) running on the Unix server (Sun, HP).

As for "the future is components, not languages," I remain skeptical. Language-independent components have been promised for several years, at least since the early days of CORBA roughly 5 years ago. Object-oriented component hierarchies have proven very difficult to coalesce into standards, because of the depth of detail that must be agreed on. I like to think that Java's JOE technology will help hasten the adoption of CORBA-based components, but we'll see....



If you have technical problems with this magazine, contact


Last updated:

Click on our Sponsors to help Support SunWorld

What did you think of this article?
-Very worth reading
-Worth reading
-Not worth reading
-Too long
-Just right
-Too short
-Too technical
-Just right
-Not technical enough

[Table of Contents]
Subscribe to SunWorld, it's free!
[Next story]
Sun's Site

[(c) Copyright  Web Publishing Inc., and IDG Communication company]

If you have technical problems with this magazine, contact

Last modified: