Congratulations on a very good issue. As someone who first encountered UNIX via Coherent and then later turned to Linux, I want to let you know that Linux Journal provided me with many interesting articles on new and useful open-source products.

As the Linux user base continues to grow, LJ is faced with addressing the needs among the growing ranks of newbies as well as those who have followed Linux for several years now. Hopefully, LJ will strike a good balance. I for one thoroughly enjoy articles on “nuts and bolts” items, such as kernel internals, libraries and shared module implementation.

Also, case studies are great. At work, we are very interested in using Linux for noncritical “ancillary” applications. These case studies help us learn from and avoid any mistakes others may have made.

Keep up the good work, and don't shy from the in-depth stuff.

PS: The “centerfold” in the June issue was more than a little tongue-in-cheek, and although I found it somewhat funny, it seemed out of place in a quasi-technical magazine. —Allan Peda apeda@interpublic.com

Today I'm ashamed of being a Linux Journal subscriber. I am disappointed with LJ and SSC almost beyond all reason. Why is the Linux Journal eCommerce site (store.linuxjournal.com) running on NT and IIS? Don't you people have a fundamental problem with that? How could you betray the Linux community in such a manner?

Is there some extenuating circumstance? Is this some sort of mistake? How little faith do you have in Linux that you would do this thing?

These are not rhetorical questions. If I don't get a satisfactory answer, my subscription is history. I read LJ front to back every month, and I would miss it terribly, but I will not support you if your politics are wrong: in a capitalist society, a dollar spent is a vote cast. Why are you voting for Microsoft? —Dale Lakes dale@multiverse.com

In order to serve our store customers better, we wanted to outsource the web presence and fulfillment to the same company. WAS offers this service, and plans to port their software to AIX. This was the best alternative we could find. If you know someone running Linux who offers the needed services, let us know —Editor

I read the article “Dynamic Class Loading for C++” in the May issue of Linux Journal. I've been experimenting with this technique awhile, and I experienced a problem that is not mentioned. I had some name clashes between libraries: if two libraries have an internal class, function or public symbol with the same name in both, once they are loaded, the executable behavior is unpredictable.

It seems that when a library is linked in, the dynamic linker starts searching all libraries for symbols, even for the libraries' internal ones. I do not know if this problem was solved in recent versions of the dynamic linker, or if there is a linker option I missed.

I solved it by putting all internal symbols in name spaces. Making sure there were no name spaces with the same name in two libraries (unless desired) eliminated the odd behavior. —Dario Mariani dmarian@fi.uba.ar

I just read the “We Talk to Everybody” article in the June issue and found it very instructive, since as a “late” Linux user, I ashamedly confess I didn't know all the names listed in it.

However, I think two more people deserve to have their names written on the “Linux Hall of Fame”: Paul McKerras and David Hinds (hope I didn't misspell their names) for PPP and PCMCIA. Without them, I wouldn't be able to send you this e-mail (I run Linux on a notebook).

I know we have to do a sort of lightweight patching on the kernel sources for these to work, but I am really thankful. —Nguyen Tuan nguyen_a@cnam.fr

I have wanted to write about this issue for quite some time now. Everybody is lamenting a lot lately about the “necessity” of having Microsoft Office for Linux as the definitive way for Linux to succeed on the desktop. I happen to have some reservations about that.

First of all, I am lucky enough to work at a UNIX-centric office, so having Linux on my workstation is a given. Aside from my primary job functions, I use Office (StarOffice, that is) for office-related stuff. I use word processors and spreadsheets equally. I used to have Windows on a separate partition as a dual-boot machine, but it's gone now. I needed the space for something else.

The office suite problem still exists for me, though. Being bilingual, I often need to use Russian fonts/encodings on a variety of machines, mostly Linux (x86 and PPC) and Macintosh. Well, it also has to be readable on Windows. MS Word has its own most weird (incompatible) way of encoding (I'd say scrambling) Russian characters, rendering them unreadable on any other hardware/software configuration. So it's unusable for me, period. StarOffice is actually very good in creating nice portable bilingual documents that can be imported into practically any other application. Bummer there's no StarOffice for Macintosh! Corel WordPerfect has the same limitations as StarOffice, but it is even further restricted regarding the OS (no Solaris version). The only “word processor” that is actually available for any platform/OS and works with any national character set is Netscape Communicator. The output of this application beats all competition in terms of universality and portability, but it does suck as a word processor.

Bottom line: I need an office suite that is reasonably full-featured (like StarOffice), available for all major and minor OS platforms (like Netscape), produces absolutely portable text output (again, like Netscape) and doesn't cost an arm and a leg (like Netscape and StarOffice). I want to reiterate that incompatible features of MS Office make it unacceptable for me even for free. Does anybody use a non-stolen version of MS Office anyway? —Alex Aitouganov Alex.Aitouganov@Eclipsys.com

I have enjoyed your guide to the Linux desktop series immensely, and I was delighted to see that you wrote about my favorite window manager, Window Maker. Of course, I wasn't too terribly happy with your lack of praise for my darling, but still, I don't want to start a holy/flame war. I merely wanted to inform you of one thing. —Robert Wade robertwade@rocketmail.com

I'm glad you liked the series. There is one more coming, which should be out in the next couple of months, that does basic coverage of the remainder of the window managers out there. —Michael J. Hammel mjhammel@graphics-muse.org

I know that exaggerating about how long one has been using Linux is a long-standing tradition for some people in the community, but it's getting a little out of hand.

In the June 2000 issue on page 98, the article claims that H. J. Lu discovered Linux in 1990. That's simply amazing, since Linus announced the project on July 3, 1991. (www.li.org/li/linuxhistory.shtml)

Perhaps Mr. Lu has the capability of seeing into the future? —Dave Whitinger dave@whitinger.net

I neglected to mention an important point in my article on syslogd in the July issue of Linux Journal: If you plan to use the network features of syslogd, it is very important that this port be behind a strong and properly configured firewall router. If not, the feature should not be used. The network syslogd interface is a great way to do remote denial of service attacks, and even some far more insidious attacks. The denial of service can be to both CPU and disk, since sending tons of spurious messages can make syslogd fill a disk drive. Moreover, such messages may be used to obscure other evil activities which you might lose in a flood of phony messages.

Even more frightening is the possibility that you are watching some log files with, say, poorly written Perl scripts that interpolate variables containing strings from syslog-generated log files. There are some clever Perl hacks where you can compose strings that will execute programs as the user running the script. Such strings could be embedded in fake messages sent to your syslog d<\#230;>mon.

The upshot of all this is that while collecting logs from multiple machines over a network is a very nice and useful feature, you must be sure to prevent machines outside your network from reaching this port.

This is an important point, and I hope you can find room for my letter! Thanks a lot!

Here is the corrected Listing 1 to include with my other comment.

Due to an accident of typesetting, the syslog.conf listing in the July 2000 issue was incorrect. The correct listing appears below.

Listing 1

—Michael A. Schwarz mschwarz@sherbtel.net