Questions and Answers
Bjorn Satdeva
The LISA VIII System Administration Conference took
place in San Diego,
CA, September 19 to 24th. The size of this year's conference
was about
the same as last year's, roughly 1200 people, which
is a lot of system
administrators in one small place! The technical content
of the conference
this year was a bit disappointing to me, because several
of the papers
addressed topics that had been covered at earlier LISA
conferences
and some of the talks were essentially rehashes of material
the speakers
had presented before.
Some of the papers I found most useful and interesting
were: "Kernel
Mucking in Top," by William LeFebvre, from Argonne
National Laboratory,
a much overdue paper describing some of the issues in
the very popular
top program, a ps substitute; and "Managing the
Ever-Growing To Do List," by Emy Edward, Northwestern
University,
a paper describing a request-like program, to manage
the admin's
job queue.
However, the main reason for going was, as always, to
touch base with
many other system administrators, and simply to compare
notes on a
person-to-person basis. So even if the speculations
voiced by some
people -- that LISA may be at a place where it is growing
ever
larger but its technical content is starting to slip
-- turns out
to be correct, LISA will still be one of the major conference
on my
agenda for years to come.
The Interop tradeshow has also taken place, this year
in Atlanta,
GA. This trade show is also growing bigger each time,
and it is in
fact getting harder to get good technical information
there. For many
of the vendors, marketing is increasingly taking precedence
at the
cost of technical information. The show can be a good
opportunity
to learn specifics from vendors chosen in advance, but,
as always,
the most fun places to go were the small startup companies,
with new
products.
Of products in the latter category, I liked the "Top
of Mind"
help disk software from the Molley Group, Inc. The product
is intended
to allow the user to interact directly with a Windows-based
program
which will ask questions and guide the user to a solution.
The product
currently runs only under Windows, but UNIX versions
are in the works.
The software demonstration was very impressive, but
I of course do
not know how well this product will stand up in a real
environment.
The Molley Group can be contacted at (201) 884-2040.
I also took the opportunity to stop at the Novell booth,
to ask about
their vision of what is to come for UNIX. So far, Novell
has totally
failed to establish any direction for the future of
UNIX, which is
in many ways unfortunate. However, I was assured by
the people at
the booth that Novell is about to change this picture,
and that we
should see evidence of this over the next few months.
It is now a
question whether this is at all possible, since Sun
has purchased
a license-free copy of the System V Release 4 sources
for future development
of their version of UNIX, and BSDI has another version,
based on BSD
Net2 and 4.4, which is free of any license requirement
from Novell
and USL.
Next, a number of upcoming events. While LISA is over
for this year,
the other main conference for system administrators
is ramping up
on the East Coast. The Fourth Conference on System Administration,
Security, and Networking (SANS IV) will take place April
24-29, 1995,
in Washington, DC. However, if you want to present a
paper at that
conference, you need to get your proposal in right away,
as the deadline
for this is November 1. Please contact sans@fedunix.org
for
further information on this conference.
Another upcoming conference is ETHICOMP95, the International
Conference
on the Ethical Issues of Using Information Technology.
It is organized
by De Montfort University, in the United Kindom, and
Southern Connecticut
State University, in the United States. The conference
is scheduled
to take place at De Montfort University, Leicester,
March 28 to 30,
1995. From the Call for Papers (the deadline for which
precedes the
publication of this issue of Sys Admin), it appears
that ETHICOMP95
will provide an excellent forum for stimulating debate
on fundamental
issues relating to the development and use of Information
Technology
and Information Systems. There will be an opportunity
to consider
approaches based on the different cultures/countries
of both conference
presenters and conference delegates. The three-day conference
consists
of three parallel themes. Each theme will comprise a
series of papers
and workshops. For more information, contact Simon Rogerson,
Co-Director,
ETHICOMP95, Department of Information Systems, De Montfort
University;
voice: +44 533 577475; fax: +44 533 541891; or email:
srog@dmu.ac.uk.
Yet another conference of interest for UNIX administrators
is the
"UNIX & The Law" Symposium, sponsored
by the Sun User Group,
in Austin, TX, to take place November 14-17, 1994. This
conference
is meant to create a forum where the members of the
once distinct
areas of technology, legislation, and law enforcement
can meet to
share experiences and ideas. The conference consists
of a day of
tutorials, two days of talks, and yet another day of
tutorials. For
further conference information, contact the Sun User
Group, voice:
(617) 232-0514; fax: (617) 232-1347; email: conference@sug.org
And now to the real question and answers.
At my site, if a user sends any internal mail from
some
machines (say group A), the mail header consists of
<username>@machine_name;
from other machines (say group B), if the same user
sends internal
mail, the header is <username> only. We have tried
copying
/etc/sendmail.cf from group B to group A machines and
have
tried sending mail after killing and reinvoking /usr/lib/sendmail,
but the result is same.
It sounds like entires in your /etc/hosts file
are incorrect for some of your hosts. The sequence in
that file is
more important than many man-pages and UNIX admin books
may
lead you to believe. The first entry is of course the
IP address for
the machine, and the second is the "real"
name of the machine,
which must be fully qualified. Following the real name
can be as many
aliases as you need. Second, you must ensure that when
you set the
machine name (typically in /etc/rc on BSD systems, and
something
like /etc/rc1.d/tcp on System V), you use the fully
qualified
domain name. If you don't follow these rules stringently,
you will
get the kind of unpredictability you described above.
Figure 1 shows
some sample /etc/hosts entries.
Do you happen to know of any products that do password
encryption across the network so that the password could
not be read
by an analyzer/sniffer? I cannot find one anyplace,
and it seems I've
asked everyone who knows how to spell UNIX if they've
heard of such
a thing.
I believe that there is no direct solution to this
problem.
I have, however, two possible solutions for you, although
they do
not address your question directly. One solution is
to implement Kerberos
everywhere, in which case the need for sending the password
over the
wire is eliminated. Another possibility is to implement
a challenge/response
system, such as the S/Key free-ware or the Secure ID
from Security
Dynamics. However, if your users do a lot of remote
logins, it will
be a pain in the neck to use the challenge response
system.
In the old days, the UNIX man pages came with
a permutated index, which was helpful if you were looking
for a specific
kind of operation or command, but did not know which
man page
covered it. The UNIX system at my site does not have
this, and to
make the situation worse, we have only the preformatted
man-pages,
so the old macros do not allow us to build an index
ourselves. Is
there any way we can get a permutated index for our
version of UNIX?
It is very unfortunate, but it seems that permutated
indexes are a thing of the past. For people who are
unfamiliar with
the permutated index, this was an index created by a
program, which
extracted the the one-line description from the top
of the man page,
and then did a sorted permutation, somewhat like this
short example,
using the man-page for man:
display the on-line manual pages man(1)
display the on-line manual pages ............ man(1)
display the on-line manual pages .................... man(1)
display the on-line manual pages ........................... man(1)
Such a permutated index is, as stated in the question
above, very useful when you are looking for information
and do not
know exactly which man page has it. We use BSDI on all
our
machines, and the number of new programs that were picked
up from
the Internet and are now included on their distribution
cd-rom is
astonishingly large. A permutated index would have been
very useful,
but BSDI did not provide one. I therefore wrote the
perl program in
Listing 1. The program traverses the man-page directory
trees,
and makes a best attempt at building a permutated index.
As it uses
the preformatted man pages, it makes some mistakes here
and
there, but still the output is good enough to be very
useful.
About the Author
Bjorn Satdeva is the president of /sys/admin, inc.,
a consulting
firm which specializes in large installation system
administration.
Bjorn is also co-founder and former president of Bay-LISA,
a San Francisco
Bay Area user's group for system administrators of large
sites. Bjorn
can be contacted at /sys/admin, inc., 2787 Moorpark
Ave., San Jose,
CA 95128; electronically at bjorn@sysadmin.com; or by
phone
at (408) 241-3111.
|