I just read coverage of a panel of programming luminaries on Salon; the topic of discussion was about the state of programming. In the course of the discussion, the subject of Open Source came up. Several of the luminaries — which included architects of the Mac OS and Windows, as well as others — derided the community for basically reinventing the wheel, and wheels that need to be re-thought entirely anyways. One questioned, "Why is the idealism just about how the code is shared — what about idealism about the code itself?"

Andy Hertzfeld (who helped develop the original Mac OS) was sitting on the panel, and jumped in. He has been working with Eazel and Chandler in recent years, and thus has an inside view of open source. His initial comment: "It's because they want people to use the stuff!" Basically, they program Windows- or Mac-like interfaces because then people will be willing to try it out. They program office suites because people "need" an office suite to be productive. Such offerings hook them into the OSS movement.

Another participant, Dan Bricklin (of VisiCalc, a pioneering spreadsheet program) shared an anecdote from Bill Gates. Evidently, Gates gave an interview (with Lammers — look up this person) in which he explained that his work on MS's BASIC compiler was done by looking at how other programmers had accomplished the task. In his own words, "The best way to prepare is to write programs, and to study great programs that other people have written. In my case, I went to the garbage cans at the Computer Science Center and I fished out listings of their operating systems."

So basically, Gates was an early adopter of OSS methodologies… Interesting to see that today he's so protective of MS code. Guess money might do that to you.

Last night, I created my first RAID array from the commandline. It was quite simple, I discovered.

1. Create your partitions using fstab. Remember, primary partitions must be created before extended partitions.
2. Look in /proc/partions and note the new partition IDs.
3. Edit /etc/raidtab and create a new RAID array. If unsure of the syntax, look up the Linux Software RAID HOWTO for more details.
4. Type mkraid /dev/md?, where ? is the id of the RAID device you just entered in /etc/raidtab.
5. Format the new RAID device with your favorite filesystem, assign it a mount point, and start using it!

I was impressed with how easy it was; the choices that the Anaconda installer present for creating a RAID array made it seem like the underlying process must be difficult, when in fact it may have been almost the same complexity if not easier.

Linux.com has had a running series on CLI commands for Newbies. Most of it has been very basic, but there are still a few gems within. For instance, today I was introduced to apropos and whatis. Give a search term to the former, and it will list all programs in which the search term is found in the manpages; give a program name to the latter, and it will tell you which man page addresses it.

I've had occasion to need to grab a specific set of files from a large directory — most recently, I needed to grab some specific access logs from our Apache logfiles at work.

Enter find.

I needed to get all files newer than a specific date, and with the pattern 'sitename-access_log.timestamp.gz'. I then needed to tar up these files and grab them for processing. So, here's what I did:

• The -newer filename tells find to locate files newer than filename.
• The -regex flag tells find to locate files matching the regular expression. The regex that find uses is a little strange, however, and didn't follow many conventions I know; for one thing, it's assumed that the pattern you write will match against the entire string, and not just a portion of it. What I ended up using was -regex '.*access_log.*gz', and that worked.
• The -printf flag tells find to format the printing. This is useful when using the output of find in another program. For instance, tar likes a list of filenames… so I used -printf "%p ", which separated each filename with a space.

I then backticked my full find statement and used it as the final argument to a tar command; voila! instant tar file with the files I need!

I just ran into this not long ago, and wish I'd discovered it years ago. Basically, Ctrl-S suspends a process, while Ctrl-Q resumes it. This is useful when in g/vim or screen and you manage to lock up your application because you accidently hit Ctrl-S when reaching for another key combo.

I wrote up a short tutorial today on the IT wiki about SSH tunneling. What I didn't know is that you can start a tunnel after you've already ssh'd to another machine. Basically, you:

• Press Enter
• Type ~C

and you're at an ssh> prompt. From there, you can issue the tunnel command of your choice: -R7111:localhost:22, for instance.

I was just reading an article about the Dean campaign's IT infrastructure, and there's an interesting quote from their IT manager, Harish Rao:

"I believe in three principles", he said. "First I always make sure I hire people I can trust 100%. Second, I always try to hire people who are smarter than I am. Third, I give them the independence to do as they see fit as long as they communicate about it to their other team members. We've had a lot of growing pains, a lot of issues; but we've been able to deal with them because we have a high level of trust, skill and communication."

I know for myself that when I (1) don't feel trusted, and/or (2) am not given independence to do what I see as necessary to do my job, I don't communicate with my superiors about my actions, and I also get lazy about my job because I don't feel my work is valued.

Fortunately, I feel that in my current work situation, my employers followed the same principles as Rao, and I've felt more productive and appreciated than I've felt in any previous job.

One thing I noticed today when using RCS is that it isn't terribly user friendly — you need to checkout a file to make edits. Often, I make edits, and then want to commit my changes.

So I wrote a wrapper script called revise. It makes a temporary copy of the file you've been editing, checks it out of RCS with locking, makes it writeable, moves the temporary copy to the permanent name, checks it in and unlocks it (which prompts for a log message), and then makes the file writeable for the user and group again. The script is outlined here:

###!/bin/bash
FILE=$1
cp $FILE $FILE.new
co -l $FILE
chmod u+w $FILE
mv $FILE.new $FILE
ci -u $FILE
chmod ug+w $FILE

Being the ROX-Filer centric person I am, I also wrote a quick perl script called rox-revise that I can then put in my SendTo menu. It parses the file's path, changes to that directory, and then calls the revise script on the filename, from within a terminal. This script follows:

###!/usr/bin/perl -w
use strict;

use vars qw/$path $file $TERMCMD $REVISE $ZENITY/;

### Configurable variables
$TERMCMD = "myTerm";    # What terminal command to use; must be xterm compliant
$REVISE  = "revise";    # What command to use to revise (i.e. rcs ci) the file
$ZENITY  = "zenity";    # The zenity or dialog or xdialog command to use

### Grab the filename from the command line
$path = shift;
$file = $path;

### If no file given, raise a dialog and quit
if (!$path || ($path eq '')) {
    system(
        $ZENITY, 
        '--title=Error', 
        '--warning', 
        "--text=No path given to $0; rox-revise quit!"
    );
    exit 0;
}

### Get the path to the file and switch to that directory
if ($path =~ m#/#) {
    $path =~ s#^(.*)/.*?$#$1#;
    if ($path !~ m#^/#) { $path = "./$path"; }
    chdir $path or die "$path not found!n";
} else {
### Or else assume we're in the current directory
    $path = './';
}

### Get the filename
$file =~ s#^.*/(.*?)$#$1#;

### Execute the revise statement
my $failure = system($TERMCMD, '-e', $REVISE, $file);
if ($failure) {
    # on failure, raise a dialog
    system(
        $ZENITY, 
        '--title=Error', 
        '--warning', 
        "--text=Unable to revise $file"
    );
}

1;

Now I just need to check out Subversion, and I can have some robust versioning!