Ten Questions For Majordomo*
(An Interview With D. Brent Chapman)
By
Jason Levitt
ailing lists have been around since the dawn of Internet time (A.D. 1985) and, like cockroaches and Elvis, they will likely persist long after I'm gone. They are resistant to spam, have low maintenance requirements, and are easy to deploy in the enterprise. Corporate users have, in fact, embraced mailing lists in a big way. Technical support departments find that mailing lists are a great way to keep developers and other interested third parties aware of late-breaking news. Marketing and sales folks like
the way
announcement mailing lists can quickly notify the public of events.
Finally, mailing lists are an easy way to handle the broadcast of corporate communications, especially in heterogenous intranets where E-mail may be the most widely used communications technology.
Talking To The Experts
Yes, mailing lists are cool, but given all the interesting Internet technologies out there such as Net news, push technology, conferencing, and the Web, it's hard to get a feel for where mailing lists fit in and where they might be heading technically. To get a better feel for the future of mailing lists, I enlisted the aid of a mailing-list expert. D. Brent Chapman wrote one of the most popular mailing list packages ever. He called it
Majordomo
which is from the Latin major domus, which means "master of the house."
Majordomo is a robust freeware** mailing-list package written using the Perl language. Many people (m
yself included), feel that Majordomo has become the standard by which other mailing-list packages are judged.
I was fortunate to catch up with the creator of Majordomo at
the 11th annual USENIX LISA (Large Installation Systems
Administration) conference in San Diego. Perhaps ironically,
it was at the sixth LISA conference in 1992 that he
presented his
original paper about Majordomo
.
D. Brent Chapman
is the owner and technical head of Great
Circle Associates. He created Majordomo, though, in the
grand tradition of Unix systems software, many people had a
hand in its ongoing development and deployment.
"I wrote the original version, but I'm not too heavily
involved with ongoing development any more, except as sort
of resident curmudgeon who occasionally sticks his head out
of his cave to explain why I originally did something a
certain way, or to tell folks what I think
of the latest
feature somebody has proposed to add to Majordomo," he says.
"There's a very dedicated and talented group of folks on the
Majordomo-Workers mailing list who've done all the upgrades
and enhancements to Majordomo for the last few years; the
key contributors are credited on the Majordomo Web page.
Special credit should go to the release coordinators who've
handled all the Majordomo releases for the last few years
(the 1.6x and 1.9x releases), John Rouillard and Chan
Wilson."
Thanks Brent. And now, 10 questions for you:
-
Statistically speaking, how fast is Majordomo growing in
terms of number of mailing lists and number of users? Have
you seen any slowdown? Major growth?
We don't track the number of lists or subscribers on
Majordomo servers worldwide. For the last year or so, since
the release of Version 1.94 in October '96, the Majordomo
installation process has been generating a "registration"
message that says what version of Majordomo is being
installed, what kind of hardware and software it's being
installed on, what version of Perl is being used, and so
forth; so far, in a little over 12 months, almost 11,000
distinct sites have installed Majordomo. The registration is
optional; we don't know how many sites installed the
software and didn't want to register it for some reason, but
I'd guess that's a fairly low percentage.
If I had to guess, I'd say the average number of lists per
server is probably about 10, and the average number of
subscribers per list is probably about 100. Those numbers
are completely speculative; I know of sites that use
Majordomo for hundreds of lists, and I know of Majordomo-
managed lists with thousands of subscribers.
-
Mailing lists are obviously becoming much more popular in
the corporate arena -- there are lots of commercial packages
available now. Do you have a feel for how companies are
using them?
Companies are using them in a variety of ways, for both
internal and external c
ommunication. Lots of Majordomo
servers are installed within organizations, and are used to
create internal mailing lists by topic, workgroup,
geography, or other interest. Lots of companies are using
Majordomo servers externally, as well, for both announcement
and discussion lists to support the users of their products
and services. Finally, some companies are supporting mailing
lists on particular topics as a public service, for the
(hopefully positive) publicity this generates for them; this
is kind of the Internet equivalent of companies sponsoring
sporting and cultural events.
- Mailing lists appear to be one of the few safe refuges
from the dreaded Spam. Has the rise of Spam impacted the
acceptance or development of Majordomo?
Many of the features added to the last few versions of
Majordomo have been driven by or directly related to spam,
such as subscriber-only posting (you can only post to such a
list if you're already a subscriber to it), confirmed
subscriptions (yo
u have to specifically confirm your request
to subscribe to such a list, with a "key" that Majordomo
mails back to you), and header/body "taboo" filters (which
cause a message to be held for approval by the list owner if
the message matches any patterns that the list owner has
specified). Of course, even the best software can't match a
human when it comes to sorting the legitimate messages from
the spam, and Majordomo has always supported moderated
mailing lists for those willing to invest the time required
to moderate a list and review every message before it is
distributed to the subscribers.
The arms race with the spammers never seems to end, but the
Majordomo-Workers folks (who are the ones developing new
versions and contributing new features) seem to be keeping
up.
-
Part of the attractiveness of mailing lists is that users
needn't have login accounts or direct access to the server
in order to participate. On the other hand, it can be
difficult to verify users' identities with
out some kind of
access controls. Are there old, new, or evolving techniques
to handle user verification?
Not really. There has been talk of a PGP-enabled Majordomo
for years, but I don't think anything has actually been
coded or released. There are a number of folks working on
Majordomo 2.0 right now, though I'm not too involved with
that effort. My understanding is that Majordomo 2.0 will be
much more modular than the current Majordomo, and that it
should be possible to more easily integrate things like
encryption and real authentication.
- Majordomo is often used as a kind of push technology
(e.g. announcement lists), yet you rarely hear it mentioned
in the same breath as Pointcast or Marimba. How do you think
it fits in to the whole realm of Usenet, E-mail, Web,
intelligent agents, and push software such as Netscape's
Netcaster?
Announcement mailing lists are the original push technology,
in my view, and they still have a lot of advantages over
more active tech
nology like Pointcast, Marimba, and
Netcaster. For example, all you need to participate is E-
mail; you don't have to run a special client to receive the
"push." The information being distributed is there waiting
for you; you don't have to be online and in the push client
when what you're interested gets pushed from the server. You
can keep the information by simply filing away the E-mail
messages. Mechanisms are already in place to optimize
transmission of E-mail to multiple recipients at a single
site, so that the site gets one message with multiple
recipients rather than a separate transaction for each
recipient; you don't end up wasting bandwidth to transmit
the same info over and over to multiple people at the same
site.
Even organizations pushing new push technology seem to
recognize the advantages of announcement-only mailing lists;
for example, every few weeks I get a "Netscape Product News"
announcement by E-mail direct from Netscape.
- It seems that Microsoft and Netscap
e have been slow to
implement mailing list solutions. In fact, I don't believe
that either sells a mailing list server. They haven't been
shy about competing in other markets. Any clues?
I've made the same observation, and I'm not really sure why
it is.
-
Scalability seems to be a problem with some mailing list
implementations. What are the key software design and
hardware resource issues that effect scalability of
Majordomo and mailing lists in general?
That's a big topic, with a lot of subtle details... The
biggest bottleneck that most sites face is bandwidth; they
simply don't have the network capacity to handle the load.
Sometimes CPU power, memory, or disk space is a problem.
Another bottleneck is the design of the mailing list
software itself. I designed Majordomo, for instance, to
handle tens of lists, each with hundreds of subscribers, and
tens of messages per day. Now, I know of sites (including
our own) using Majordomo for thousands of lists, I know of
sit
es using Majordomo for lists with thousands of
subscribers, and I know of sites using Majordomo for lists
which get hundreds of messages per day. You can make it work
beyond what it was designed for, but it takes some careful
analysis and configuration.
-
Have you encountered any particularly interesting uses of
Majordomo? Perhaps organizations that have hacked the source
or use it in an unconventional manner?
Oh, all the time... I recall with great pride that when I
originally released Majordomo, one of the first
organizations that I heard from who was using it was the
United Nations; they had set up a Majordomo-managed mailing
list for munitions and ordinance disposal experts from all
over the world to discuss solutions to the problem of
unexploded land mines. I'm very pleased that Majordomo has
played a part in the formation of any number of "virtual
communities" on a huge range of topics.
- In your LISA VI paper, you mentioned that you basically
started from s
cratch with Majordomo, implementing it under
Unix. If you had to write it again from scratch, with your
current knowledge of standards and cross-platform
environments, what, if anything, would you do differently?
Well, what I didn't say in that paper is that Majordomo was
also the first significant program that I ever wrote in
Perl; in the early versions, you could tell which pieces of
the package had been written in what order, because the code
got better as I became more comfortable with the language.
If I were doing it again today, I'd definitely pay more
attention to standards such as MIME and Delivery Service
Notifications. I'd also pay more attention to the problem of
bounced E-mail because of broken or outdated E-mail
addresses; that's turned out to be a huge problem that I
didn't anticipate. For example, for the lists we host here
at GreatCircle.COM, we routinely process 10 to 100 times as
many bytes of bounced E-mail every day as we do bytes of
actual submission for the li
sts. The bounces keep
threatening to bury us! I'd also create a Web-based
interface to Majordomo, at least for subscribers and list
owners. Much of what I'd do (or do differently) has been
done as add-on packages for Majordomo (for instance, Bill
Houle's "MajorCool" Web interface), and much of it is being
done by the Majordomo 2.0 developers. I'm glad to see that
folks think enough of Majordomo and what it represents to
think that it's worth improving and building on.
I still think that Perl is the right language and Unix is
the right platform. Mailing-list management is all about
slicing and dicing strings, messages, files, and so forth,
and Perl does that better than any other language out there,
including old standbys like C and C++, and flavors-of-the-
month like Java. Windows NT might someday take over the
world, but it ain't there yet; the vast majority of
significant servers on the net are still Unix-based, and I
believe that will be true for some time to come. On the
other hand, NT
is clearly carving out a niche for itself in
small-scale and departmental server applications, so I think
it would probably be worth some effort to ensure that
Majordomo ran as well under Perl on NT as it does on Unix.
- What technical direction do you see Majordomo, and
mailing lists in general, heading in the future?
Ever upwards. I think that existing lists are going to
continue to get bigger and busier, and that more focused
lists will be created in response to that. I think that
selling mailing-list management software and selling
mailing-list management services are both viable business
opportunities now, and there are already a number of
companies taking advantage of those opportunities; I see
those opportunities increasing.
As I said in my original Majordomo paper in 1992, I wrote
Majordomo for my own purposes, to solve a particular problem
that I'd volunteered to address. I'm absolutely stunned and
honored beyond description that so many other folks have
fo
und it useful for their own purposes as well.
*majordomo - n. 1. a person who speaks, makes arrangements,
or takes charge for another. 2. a steward or butler. From
the Latin major domus or "master of the house."
**Some licensing restrictions apply. See the Majordomo
license
for details.
AuthorITies Archive
Send Us Your Feedback
|
Rich Levin:
Run T
ime
Rich fills you in on all of the latest products, issues, and trends in application development.
|
|
Stuart J. Johnston:
Redmond Watch
As our eyes and ears in Redmond, Stuart gives his perspective on the latest events at Microsoft.
|
|
Charles Pelton:
Eye On I.T.
Charles explores IT management issues and strategies that business and technology managers face.
|
Top of the Page
|
