More Letters to the Editor

These letters were not printed in the magazine and appear here unedited.

Network Administration with AWK

Having written many administration tools in AWK over the years, I was pleased to see an article in your latest issue with that title. Too bad that's not what the article was about. Further a shame that every code snippet he included merely pulled in some external data and sent it to stdout. And to further suggest that one use grep to extract the data! AWK deserves so much more.

—Chip Christian, chip@princetontele.com

Feature Page

With the help of the March issue of the Linux Journal I have been able to write a feature page for the Sheffield Linux User Group site.

The authors of the original authors have written to me to help me to complete the Sheflug feature for April. I will upload the pages at the end of the month. I would like to say thank you very much to the staff at LJ. If I hadn't subscribed to the LJ I wouldn't have been able to produce such an interesting article for Sheflug.

The proposed subject for the May feature is VNC servers with a quick mention of VM Ware.

—Richard Ibbotson, r.ibbotson@zen.co.uk

Puffins and H-P

The Thursday Wall Street Journal article, “The Puffins Guide H-P Into an Alien Culture On a Software Mission”, reminds me of an elephant dancing with a mouse. I wonder which one is larger; the multi-million dollar corporation or the Linux internet culture. Time will tell.

Christopher Beard and Alex deVries were wise not to ask for any money. Just ask Ron Jones owner of Colossal Graphics and developer of a bulk ink system who sued H-P for reneging on a deal. He was awarded $6.4 million-dollars by the Hon. Joseph F. Biafore on January 15, 1999.

Sorry fellows, there is no check in the mail!

—Gerald Green, GG10MAN@aol.com

Error in April issue in “grep:searching for words” article

In this article Jan states “grep also accepts regular expressions....... the following command can be given grep flip *”.

Of course, grep does indeed process regular expressions, but the example above is not a demonstration of this ability.

It is misleading because in this example grep does not see any regular expressions. The “*” is expanded by the shell, not by grep. The shell then passes the list of filenames to the grep command.

This is an important point because newcomers to UNIX need to understand the power of the shell. Some other operating system, such as IBM OS/400, do not have facilities such as globbing built into the shell, but instead pass the responsibility to each command program.

An example of a regular expression in grep is: (a) $ grep [pw] filename
which lists all lines in the the file “filename” which contain either 'p' or 'w'.

Or you could have the same example, with shell globbing: (b) $ grep [pw] file* Which would do the same thing, but for all files starting with “file”. Say the directory holds three files named “file1”, “file2”, and “file3”, and you type in example (b) : the shell expands the command string into grep [pw] file1 file2 file3, forks a new process and executes grep. So grep doesn't even “know” about the use of “file*”.

Of course, shell globbing can be used with many different commands, not just grep. But beware, not all commands can process multiple files.

—Tim Lewis, timlewis_atlanta@yahoo.com

ARKEIA (fwd)

I read your article about ARKEIA and downloaded it. Since I don't use tape drives, but hard disk for my back up, I have some problems that should be noted. I set up everything by instructions, but it didn't work. When backup starts it says: insert tape /mnt/backup (this is where my backup disk is mounted). I say OK, but it's not satisfied. When I check all of the setup files I noticed, that ARKEIA for tape type “file” automatically asiges “manually mountable” flag. It can not be changed to “on-line”, since it is set automatically. So I tried with /dev/hdb1 as tape location, but without success. I think that a lot of readers of Linux Journal use disk's as their back-up devices. I read all the documentation (more than 300 pages), but found only one or two paragraphs dedicated to file backups. I would really appreciate it if you could try it and report it to the LJ for readers to know about it. I found information in FAQ, but then I noticed that I would need non shareware version to get it working (you need support for lib management, which is not included). I contacted ARKEIA support and they confirmed it. I told them that I find it strange limitation for the shareware version and they agreed to think about it. Support was, as you said, very quick and friendly. But facts are facts and I thing LJ readers should know about it.

Have fun!

—Bostjan Vlaovic, bostjan.vlaovic@uni-mb.si

Re: ARKEIA (fwd)

Hi, Bostjan.

I don't have the docs for Arkeia handy, so I'll comment on this from memory. Arkeia lets you define files of various sizes as “tapes”, which you then define as elements of a drivepack. I haven't tried this recently but I believe the version I used when I wrote the review would let me back up to files. If so, can you define a directory structure on your backup hard drive that lets you rotate the backups?

I'll experiment with this next time I get a chance & let you know what I find. If you get that working, please let me know.

Then there's the old standby, tar and gzip. Most backup programs compress each file individually so that they can then recover any one file later at restore time. However, if you tar then gzip, gzip compress the entire tar file as one file. This results in better compression, but may require a temporary file later at restoration time. If you have the hard drive space to do it, you might find that will do what you want.

—Charles Curley, ccurley@trib.com

Oh no . Not Another Microsoft.

Dear Linux journal Editor and Netscape team.

I would like to comment on the Letter Red Hat Phenomenon which was published in the March 199 issue. I think that Mr. Reilly is right and that you are not taking this matter seriously enough. Do we want another Microsoft in face of RedHat Linux Monopoly.

At Netscape's website I saw the advertisement of their LDAP server . http://home.netscape.com/download/noframes_102_21.html#specs Netscape directory Server was advertised to be workable on Linux.

After downloading it and going to the installation page I found out to my surprize that it works with RedHat Linux only. http://home.netscape.com/eng/server/directory/4.1/installation.html

I tried to install it on SuSE Linux 5.3 and 6.0 and nope, didnt work.

This is scary.

I dread of the day when products would say .

Installation Requirements:
Solaris 2.6
Windows NT 5
Red Hat Linux x.x

I think that Linux journal is great and is one of the very few journals that I really take time to read.

—Atif ghaffar, webmaster@artemedia.ch

letter to the editor

I would like to thank everyone who responded to my article, “Linux in a Public High School.” I would like everyone to check out my latest endeavor, the High School Linux User Group. The HS-LUG right now is just a few web pages and a mailing list, but I would like people to have somewhere to turn when they want to ask people about the uses of Free Software in education. The HS-LUG is at http://hs-lug.tux.org.

—Andrew G. Feinberg, andrew@ultraviolet.org

Linux Journal

A few notes:

1) Ethernet frames range from 64 to 1518 bytes (1514 misses the CRC).

2) CSMA/CD's “capture effect” is largely a figment—in many years consulting on thousands of networks I've yet to see it for more than some milliseconds, primarily because systems do not have resources to send at continuously full rate and a collision randomizes things quickly—yes, in some testbeds, you can demonstrate one sender's dominance for a while, but you have to work at getting the conditions right. Years ago there were chip sets (e.g., from Seeq) that violated the inter-packet gap spec, so indeed could capture a net—Cisco loved them—but the uproar was so great that Seeq decided to back off (no pun).

3) Dropping frames indeed slows down TCP a lot, due to its inability to distinguish error loss from router-congestion loss. However, this removes traffic, rather than “...leads to more congestion”. The timeout and retransmission times start at about a second and double up to about 32 seconds, whereupon transmission stops and higher layers are informed of a network error. The attached Powerpoint graph of a real TCP situation, where only errors caused loss, shows how significant delays (lower traffic) occur after each loss—blank spaces between triangles and upward drifts in bytes that could have been sent but weren't. As a protocol, TCP still needs help in these respects.

4) “...the physical distance between the furthest nodes...”—“farthest”. The editor should know that “further” refers to process (she furthered his election) and “farther” refers to physical distance.

Again, good article.

—Dr. A. B. Cannara, cannara@ibm.net

In Defense of Red Hat and other subjects!

First let me introduce myself. I am H. Aurag, a Ph.D. student at Univ of Montréal (math). There, I have been using for a while different Unix machines, from dummy terminals to SGI and Sun Workstations.

On my HOME computer, I have Windows95/RedHat Linux installed since 1998. Needless to say what I think of Windows95. The only reason I am keeping it is StarCraft, and some home software.

I think I might be the only real home user of Linux, in that I never use it elsewhere (I have no choice or say there). To tell the truth, going from Unix (Iris, Solaris) to Linux was not that hard for me as opposed to someone who would go from a Win95 only environment. But the first thing I noticed was that Linux is NOT Unix. It looks like unix, uses shells and X etc... BUT it's not Unix. And the surprise was that I found it superior, for the use I make of it (Math computations symbolic and numerical plus TeX, web surfing, and soon Civilisation CTP), than Iris or Solaris. I even found that my lousy 166MMX machine was faster than some Ultra Sparc workstations: I actually did some benchmarks, and had a big laugh.

This is the first thing I wanted to say to Unix users. You will feel comfortable with Linux, but PLEASE don't think of it as a cheap Unix clone. IT IS NOT. And I hope it will never be. Some distros might want to let you think that, but I don't agree. My hope is that Linux evolves as a unique and powerful OS (it is already) like it did by incorporating the best ideas from anywhere even NT if need be!

Also, I think Red Hat has made a great job (I really don't know any other distro, and I am not interested in knowing them). They gave me something I could use and install easily, something that makes my 166MMX with 64Meg machine faster than an Ultra5 with 96Meg and 270Mhz RISC and ALL that uses Solaris. When I am saying faster, I am talking about programs that make heavy symbolic computations and heavy floating point operations!

So kudos for RedHat, for KDE, for GNOME (I like it most! Excellent job!), for all those programmers that make nice programs and nice GUI's for newbies, also to Debian, Slackware, Suse... even if I don't know them. They all work hard I think!

Let's hope Linux keeps up with the good work and I hope I will be able to have all those games browser plugins and I hope StarOffice will realease a 5.1 version that uses less memory ;)

Finally, I would like to say to all those newbies out there, that a full RedHat install takes 40 minutes or less, and a full Win95/98 install takes 2 hours! (If you read the manuals that came with both OS'es before installing).

—H. Aurag, aurag@geocities.com

grep -c

In LJ April 1999 Take Command article on the grep command the writer states that the -c option “outputs the number of times a word exists in a file”. According to the man page the -c option “print a count of matching lines”. This difference will be apparent if the word appears more than once on the same line.

Thanks for a great magazine.

—David Massey, david.massey@ub.uio.no

Linux in Brunei

It appears that dear Stefanus Du Toit [sdt@ultracool.net] is loneley, thinking he is one of the few people who use Linux in Brunei.

Well, dear Stefanus !! Be lonely no more !!

Brunei-LUG needs people such as yourself who are willing to collaborate with other users, and to advocate the use of Linux within Brunei.

We also deal in the distribution of GPL'd Linux Distributions, please contact us for the latest distributions: Red Hat 5.9, 5.2, Debian 2.1, Slackware 3.6, Slackware 4.0 beta etc etc.

Check us out at: http://irb.lh.umist.ac.uk

oh, and if you check up the Linux counter, you might notice something interesting:
http://counter.li.org/bycountry/BN.html

Shows it's just the two of us :)

—Izam-Ryan Bahrin, izamryan@netscape.net

Minor correction and Letter to the Editor

Linux Journal is amazing, and issue #60 was no exception. However, I'd like to point out a possible minor typing mistake on article “Linux 2.2 and the Frame-Buffer Console”, page 14: under the “Frame Buffers” subtitle, on the 2nd paragraph, the text says the frame buffer device files are “/dev/fd*”. Would them be “/dev/fb*”?

On the same LJ issue, reading the Letters To The Editor section, the letter “SSC's Distribution Choice” raised an interesting point for me: the fear of having the Linux software market being directed to a specific Linux distribution. Although we know that Linux is just the OS kernel, users frequently see Linux as an entire package, from the installation program to the applications that come with a distribution. So, there are of course chances that a specific distribution be “preferred” by the general Linux user. In my opinion, when a company says it will support a specific commercial distribution, either that distribution has a special characteristic (such as glibc), or there is contribution from the distribution's company side on porting their products to Linux. I am not a RedHat or Caldera fan, but I must admit these are examples of companies that worked together with some commercial software developers to port their products to Linux. Anyway, I believe that any Linux software geared towards a specific distribution can be run on any other distribution. What is needed is some standardization between distributions, such as to include or not to include glibc, to use or not to use .RPM packages, the filesystem layout, and so on.

Sorry for a so long letter.

—Celso Kopp Webber, webber@sj.univali.rct-sc.br

Certification

I totally disagree with P. Tobin Maginnis, article “Linux Certification for the Software Professional”, April 99 Linux Journal. I am actually a little disappointed in you for publishing it. I feel that some readers who don't think too deeply about what is presented will be seduced into agreeing.

In the opening paragraph Maginnis says “ [certification] will be essential in the future as Linux software is brought into corporate and government environments.” This has not been true for any of Linux's competing o perating systems. I believe that Linux's chances of penetrating these markets will depend upon its performance and stability compared to its competition.

Instead of giving us convincing logical arguments why we need to license ourselves, he points to so many other fields that require a license as ifthis is enough reason. Next, “the U.S. Department of Labor predicts ..... 300,000 more ..... positions than programmers ..... and universities will not be able to fill the demand .....” Here he's trying to scare you and at the same time he's hoping you won't realize that such certification will limit or reduce the number of programmers that qualify which only aggravates the problem.

“..... a fundamental concern of corporate and government managers is the cost of operations—and Linux dramatically reduces this cost.” And if we will just accept certification then one of the benefits is “ ..... more job options and ..... better pay”, thus higher operating costs.

“The consensus among managers is that certification leads to improved employee service .....”. What group of managers of “certified” programmers produced this “consensus”? Assertions like this with no supporting evidence should be regarded as false.

“Licensing is a governmental action that seeks to safeguard the health, safety, welfare .....”. I say BULL! When was the last time you saw anyone go after a bad doctor, lawyer, engineer or accountant? When was the last time you saw one doctor, lawyer, engineer or accountant testify against one of his fellows? I am sorry, but, once licensing gets in place all the license agency cares about is its own bureaucratic rules, forms and fees.

“..... the expected rapid growth of Linux will create a need for business and government managers to have an additional metric in the selection of new employees .....”. Why is a new metric required? STANDARDIZING is not an answer! If you inspect every programmer ad in a major city newspaper you will NOT find two alike. Nor will you find any two employers with the same system setup or skill requirements. No licensing bureau can improve the employee interview process. Only training your managers in management and interview skills can do that.

Instead of presenting this technical forum with well organized facts and logic, Mr. Maginnis has approached this subject in a typical liberal fashion. He has used circular logic, scare tactics, contradiction, me-too ism, and unsubstantiated assertions presented as if they were facts.

Microsoft figured out that the performance of their products could be degraded by poor technical skills. This in turn, hurt the customers' impressions of their products. The “Microsoft Certified Systems Engineer” program was started to cure both of these problems and enhance their sales. But, notice, this was done “within” the industry. And, it was done their way and on their timetable with no government involvement or interference.

I will never willingly support the involvement of any government agency in the programming profession. If major players in Linux see problems similar to Microsoft, maybe they should pursue a similar solution. If you do some research, you will find that there have been previous attempts like this one and all of them have failed. Anyone remember the DPMA's Certified Data Processor?

—Richard McClendon Jr, rhmjr@flash.net

Your Article “Linux Certification” in April 1999 Linux Journal

Dear Dr. Maginnis:

As an MCSE/MCP+I/MCT, I couldn't disagree more with you regarding the “need” for Linux certification.

Every certification program I've seen through the years (A+, CNE/CNA, MCSE, etc.) starts out with the best intentions, but eventually loses its way by getting caught up in the “business of certifying people”.

One problem is that no one way of becoming “certified” is the best. One thing that “should” happen is that the companies doing the certifying need to properly screen prospective students to properly guide people with different levels of skill into the appropriate training tracks, & recommend opportunities for gaining hands-on experience so that we don't end up with “paper” MCSEs, CNEs, whatever. My experience as an MCT working with ATECs as a contractor is that they don't do this of course because that path of integrity could cost them money.

As an MCT it amazes me the expectations people have who get sold these certification programs by some sales rep working for an ATEC. They think that last week they were an interior decorator, they went a “Bill Gates Wants You” career night, now they're in this certification program, and in a couple of months with their freshly minted MCSE they'll get $60,000/year starting salary without any experience. Especially pernicious in my mind in this area are the so-called “boot camps” which cram certification preparation into a couple of weeks.

Another issue is that just because a job candidate has an acronym following their name doesn't mean they have the requisite skill level; employers still have a responsibility to screen these people to avoid hiring someone unqualified.

When I was a contractor managing a web development project last year, I consistently told my company to require all new hires to agree to work on 90 day contract-for-hire to prove themselves.

Another problem with cranking out a certification program is the logistics of developing and deploying quality materials. I am phasing out the MCT portion of my training business and focusing on custom training and IBT development, actively turning down offers to teach at ATECs primarily because of having to use Microsoft's Official Curriculum materials, which are of “inconsistent” quality.

Based on my experiences, my advice to the Linux community regarding certification is if you're going to do it, learn from the mistakes of your predecessors and don't let the process turn into a “puppy farm.”

Sincerely,

—R Chance Brents, brentorg@home.com

Re: Your Article “Linux Certification” in April 1999 Linux Journal

Dear Mr. Brents,

Thank you for taking the time to comment on my article in the Linux Journal. I must say that I am impressed by the depth of your comments, your knowledge of the industry, and your true concern for quality education. Allow me interleave my other comments with yours.

As an MCSE/MCP+I/MCT, I couldn't disagree more with you regarding the “need” for Linux certification.

Every certification program I've seen through the years (A+, CNE/CNA, MCSE, etc.) starts out with the best intentions, but eventually loses its way by getting caught up in the “business of certifying people”.

One problem is that no one way of becoming “certified” is the best.

One thing that “should” happen is that the companies doing the certifying need to properly screen prospective students to properly guide people with different levels of skill into the appropriate training tracks, & recommend opportunities for gaining hands-on experience so that we don't end up with “paper” MCSEs, CNEs, whatever. My experience as an MCT working with ATECs as a contractor is that they don't do this of course because that path of integrity could cost them money.

But at the same time, these training centers will only teach what they are allowed to teach by the vendor. So it seems that you are saying that the vendors are not providing the proper training materials? Or maybe you are saying that the vendors are just throwing out a static product and not receiving information on the type of student studying the material.

Tell me if I am wrong, but it seems to me that the more material the vendor offers the more opportunity the trainer has to train. So the trainers would welcome more “tracks” to teach.

As an MCT it amazes me the expectations people have who get sold these certification programs by some sales rep working for an ATEC. They think that last week they were an interior decorator, they went a “Bill Gates Wants You” career night, now they're in this certification program, and in a couple of months with their freshly minted MCSE they'll get $60,000/year starting salary without any experience.

Especially pernicious in my mind in this area are the so-called “boot camps” which cram certification preparation into a couple of weeks.

Another issue is that just because a job candidate has an acronym following their name doesn't mean they have the requisite skill level; employers still have a responsibility to screen these people to avoid hiring someone unqualified.

When I was a contractor managing a web development project last year, I consistently told my company to require all new hires to agree to work on 90 day contract-for-hire to prove themselves.

Another problem with cranking out a certification program is the logistics of developing and deploying quality materials. I am phasing out the MCT portion of my training business and focusing on custom training and IBT development, actively turning down offers to teach at ATECs primarily because of having to use Microsoft's Official Curriculum materials, which are of “inconsistent” quality.

Based on my experiences, my advice to the Linux community regarding certification is if you're going to do it, learn from the mistakes of your predecessors and don't let the process turn into a “puppy farm.”

Sincerely,

—Tobin Maginnis, ptm@sairinc.com

Internationalization

Recently I tried to fax a letter to your subscription department, and had difficulty dialling the number given in the contact information at the front of the journal. I dialled my international carrier, then the country code (281), then the number. Then I thought to myself “Hmm, 281 is a strange country code”, and it dawned on me - it's an area code ! The country code is 1, of course. What else could it be ?

I think that the failure to list any of your contact numbers in international format is embarrasing. A great deal of the Linux community is outside the US, and indeed the origins of Linux itself is from outside the US. FYI, the format for international listing is +country-area-number or (+country) area-number. Given the theme of the March issue, don't you think you should get with the program ? :-)

Cheers,

Brian Lowe, lowe@mds.rmit.edu.au