Mick—thanks for your Paranoid Penguin columns; I always open my new issue of LJ right to it to see what you've got in store each month. The “Hardening Sendmail” article [LJ April 2002] was another winner. I just wanted to point something out regarding your recommendation not to run a sendmail dæmon if only sending mail. I used to do just that but found something unfortunate: if a send attempt fails, then sendmail could queue it for later delivery. If the sendmail dæmon wasn't running, well, “later” would never come. I wouldn't know about the failed delivery (especially without the usual informative messages from sendmail after four hours, and after five days). So, I run sendmail (postfix, actually). If you've got a better solution, I'd love to hear it. And I'm sure other readers would, as well.
Mick replies: Thanks for the kind words! You're right, if sendmail isn't running as a dæmon, queued mail will remain queued indefinitely. I should have mentioned that common practice is to set up sendmail to be run from cron periodically with the -q flag. For example, this sample crontab line (adapted from one in Olaf Kirch's Linux Network Administrator's Guide) invokes sendmail in “flush queue” mode every 15 minutes:
0,15,30,45 * * * * /usr/sbin/sendmail -q
Obviously, that has to be in the crontab of an account authorized to run sendmail this way—usually root. And depending on how much outbound mail you deal in, you may not need to run sendmail every 15 minutes—hourly may suffice. Postfix most definitely rocks, by the way. And my friends who use qmail *really* like qmail.
I was surprised to see your article about P3P in the April 2002 issue of Linux Journal. Since 1999, P3P has been shown to be orthagonal to privacy, despite claims made to the contrary. I don't really see that the W3C has done a lot to allay fears of P3P. P3P is overly complicated and is geared toward collection of user information, not protection of it. The only serious difference between P3P and non-P3P sites is convenience in giving away your personal information. It's not even a legal help: just because a company does something illegal does not mean the average Joe can do anything about it.
The protocol could be greatly simplified and need not have any information about the user at all. Also, even if the protocol forces contact information to be given to the user, there is no easy method for the browser to determine its validity, and it is no guarantee the company will listen. In other words, “same ole same ole”, but more complex. A protocol that really is designed to protect users' privacy will never need to know anything about users except their privacy preferences. There is absolutely no need for other information, yet P3P includes a large amount of detailed and personal user information. You have to ask yourself why this has been made so complex and so heavily geared toward data acquisition.
I have been a Linux user since Red Hat 4.2 and picked up my first LJ about two years ago. I have always poked about in all the different directories and have always wondered “What is really required here?” I mean, what is the minimum requirement for a working Linux distribution, without any user apps, just something that loads, prompts for a login, then takes you to a shell prompt and lets you log off or shut down. No lynx, no elm, no sendmail, no anything. I have always thought it would be a great personal educational undertaking for myself to attempt to create my own distribution...but I've no idea where to begin.
Yes, you can build a tiny distribution with just a kernel and a shell. You might want to start with Brian Finley's “Brian's Own Embedded Linux” and remove software from it. BOEL fits on a floppy, and two articles about it appear on embedded.linuxjournal.com. See www.linuxdevices.com/articles/AT9049109449.html and www.linuxdevices.com/articles/AT5974781081.html.
A note in the UpFront section of the May 2002 issue describes the Fórum Internacional do Software Livre as an “open-source event” and says that the state of Rio Grande do Sul promotes and leads in use of “open source”. Actually the event is about free software (software livre, in Portuguese), and the state's policy is to promote free software.
I launched the Free Software movement in 1984 to campaign for computer users' freedom to have a community. To make freedom and community possible, we began developing the free software GNU operating system, which is the basis of the GNU/Linux system that your magazine is dedicated to. Our idealism made our community possible.
The Open Source movement has contributed to our community, and its supporters have a right to promote its views, but we are not part of them. The Porto Alegre event's organizers and the government of Rio Grande do Sul know the difference between the two movements, and they chose the Portuguese term for “free software” to say where they stand. I ask that you not label Free Software movement supporters, our work, our events or the community we built, with the term of the other movement. For more explanation of the difference, see www.gnu.org/philosophy/free-software-for-freedom.html.
One other correction is that I will not be speaking at this event. It was originally planned that I would, but I visited Porto Alegre for the Fórum Social Mundial in January 2002, so we decided it would be more interesting for Bob Chassell, cofounder of the Free Software Foundation in 1985, to go this May instead of me.
—Richard StallmanPresident, Free Software FoundationChief GNUisance of the GNU Project
In his article “Intrusion Detection Systems”, Pedro Bueno proposes a sample Snort rule that is supposed to alert us of “any porn web access attempt” from the private network. Unfortunately, in its current form, the rule will trigger only when a user is posting data containing the string “free porn” since it only filters outgoing traffic. In order to detect actual access attempts, the rule should filter the incoming traffic:
alert tcp $EXTERNAL_NET 80 -> $INTERNAL_NET any (msg: "Web Porn Access Attempt"; content:"Free porn"; nocase; flags:A+);
An excellent article, though.
On page 14 of the May 2002 issue, in the “It's Trivial” section, question seven, the title of William Gibson's novel should read Neuromancer rather than Necromancer. Sincere thanks to the many Gibson fans for pointing out this unfortunate typo.
Due to a printer error, page 20 of the June 2002 issue appears twice, leaving out page 21. Please see the full article on our web site: www.linuxjournal.com/article/5839.