Letters to the editor

Corrections to Dan Shearer's summary of qmail

Greetings,
 
In your most recent issue, you published the first part of an article by Dan
Shearer about the various common Unix-based MTAs. Unfortunately, his summary
of qmail contains some fairly glaring errors (both of fact, and of more
debatable issues). As a long-time member of the qmail-using and -supporting
community, I feel I'm in a position to help correct the record.
 
Errors of fact in his article include:
 
1) Wrong website. He gives the URL http://www.qmail.org/ as the website of
qmail. That's not the case; D.J. Bernstein (qmail's author) maintains a
website for qmail at http://cr.yp.to/qmail.html . The qmail.org site, while
very useful for qmail users and administrators, is not officially affiliated
with qmail; it is a community-based site run by Russell Nelson.
 
2) Wrong release count and dates. He claims qmail was last updated by the
author in 1997. qmail 1.03 was released in June of 1998, as is clearly listed
on the website and in the software documentation.
 
3) Wrong information about contributors. He claims there have been no major
contributors to qmail other than its primary author. That's not the case; a
simple perusal of the included documentation reveals a number of additional
contributors. Fairly major portions of the code were based on some users
contributions, including the included POP3 server and authentication framework
for such (based on code contributed by Russell Nelson).
 
Other erroneous or misleading statements he makes are more debatable. For
instance, he says qmail's source is "usable within very tight restrictions".
On the contrary, qmail's author explicitly states that the source code and be
modified and used at will for any purpose; the only restrictions the author
places on it are on redistribution -- there are absolutely no restrictions on
its use.
 
Another statement seems like trolling: "No, qmail isn't a realistic option
these days". That would come as a surprise to the millions (literally; see
the SMTP surveys) of sites running qmail. He also states "it isn't possible
for someone else to maintain it", which also comes as a surprise to those of
us who *do* maintain it; see http://qmail.org/netqmail/ , where a group of
regular qmail users maintains the legally-distributable, currently-maintained
version. This is particularly galling because he goes on to actually refer to
netqmail in the next paragraph.
 
Charles Cazabon
 
P.S. Note I am an LWN subscriber, but I post from another address because of
the volume of spam a posting on the LWN letters page attracts.
 
 
--
-----------------------------------------------------------------------
Charles Cazabon <web-feedback-lwn.net-letters-danshearer-98.23.19.82.34.97.82.34.8@discworld.dyndns.org>
-----------------------------------------------------------------------

Comments (3 posted)

The New Dependency Hell

Dear Editor,
 
I felt recently in the situation of installing an email server.
Because the system you know best and have built your remaining
infrastructure like PCs and Servers on is the most secure system - I
went with CentOS.
CentOS is a 100% spin off of some american red thingy distribution
(that company requested to remove all hints that they are related in
some sort). I chose "minimal installation" which presented me with
around 700MB of installed packages (very big for minimal I'd say).
After removing each and every unneded package I came down to ~400MB.
 
Because the configuration files have long gone beyond hand-editable
(only replacing the hostname requires touching a handful of files
because it's referenced more than once like in DHCP_HOSTNAME and the
network profiles etc) I decided to go with the natural tool
"system-config-network".
 
#> yum install system-config-network
 
which is about 397K (pretty big, isn't it).
But the dependencies weighted in at no less than 40 packages ranging
from Corba implementations (ORBit2) to gtk2 to even alsa-lib,
audiofile and esound (sound is surely needed on a server) to even half
of the gnome-libs.
 
And now for the best:
Why I even need an OpenGL library (xorg-x11-Mesa-libGL) to *JUST*
configure my network properly is very beyond my understanding. I'm an
aspiring PhD with my main field in computer science and I do care
about code reuse and centralized functionality, but this is beyond
ridiculous. The KISS principle has been violated so ugly you even
can't hear it screaming.
 
If any of the red thingy distribution employees reads this: please put
back the administration fun into your distribution!
And please also update the text utilities to cope with the current
configuration policies (netconf doesn't, as doesn't
system-config-network-tui).
 
Thank you from an admin who has been active for 11 years in the Linux field.

Comments (23 posted)

Page editor: Jonathan Corbet