Robert Pogson

Dana Blankenhorn knows a thing or two about IT but this is just wrong:
“The chasm is that point in the “s” curve where a product is known and liked by experts or aficionados, poised on the brink of mass market success or failure. Most products fall in. Why do some cross and others don’t? That’s Moore’s study.

Important examples are to be found in the world of open source. Linux crossed the chasm on servers. It failed to do so on desktops. Yet Android, a Google-developed Linux distro, seems fated to succeed.”

He presumes GNU/Linux has failed to cross some imaginary barrier when that is not the case. GNU/Linux on the desktop has long ago ceased to be a geek/early-adopter thing. It is being accepted on the mainstream/mom and pop desktop now. That started happening with the netbook. Many millions of netbook users are not geeks and don’t know an OS from an application. They know GNU/Linux works and love it. OEMs have passed up that opportunity in some ways but consumers have not. They have bought GNU/Linux whenever and wherever it has been offered. The news that Lenovo sold a million Ubuntu boxes last year was a surprise to many. How many geeks are there, 1% of users or maybe a few %? Geeks might buy 3 million PCs a year globally but tens of millions of GNU/Linux boxes are being sold, way more than geeks can absorb.

Further, there is all kinds of evidence that consumption of GNU/Linux is increasing:

• adoption in schools, and governments (not early adopters, for sure)
• huge numbers of mirrors with high volumes needed to ship distros
• Ubuntu breaking even and entering expanding lines of business
• continued growth for RedHat and Suse and it’s not all on the server
• mirror stats showing 150 megabits/s – 450 active requests average e.g. http://csclub.uwaterloo.ca/stats

There are hundreds of mirrors for Debian GNU/Linux around the globe. FLOSS mirrors put a major load on the web. One mirror might service hundreds of thousands of clients in a year meaning more than 10 million clients run GNU/Linux. Then there are places that have internal mirrors we do not see.

GNU/Linux crossed the desktop chasm years ago, probably 2007 was the point, when ASUS put it on the eeePC. I think 2009 was the year everyone except Dana Blankenhorn and amicus_curious caught onto the phenomenon.

Oracle and the properties it purchased with SUN Microsystems are evolving. OpenOffice.org is being replaced by LibreOffice and the Document Foundation. No word yet on whether or not Oracle will be participating. They let OpenSolaris wither. They may do the same with OpenOffice.org.

see http://www.documentfoundation.org/

There may be a dull year or two in FLOSS but no dull decades… Growth and change happen.

Try out the fork here: http://www.documentfoundation.org/download/

There are issues as can be expected: en-US only for now etc.

Debian Squeeze has OpenOffice.org3. I doubt there is any way the new fork will be ready or accepted in time for the release.

OpenOffice.org has been the killer app for GNU/Linux in schools. The savings on licences for Office alone can justify migrating to GNU/Linux. As Oracle seem not particularly friendly/responsive to FLOSS, this fork may have been inevitable or just a good thing on its own. This action may briefly distrub development but I see a bright future for the software in GNU/Linux.

UPDATE Here is a comment by Mark Shuttleworth.
“Office productivity software is a critical component of the free software desktop, and the Ubuntu Project will be pleased to ship LibreOffice from The Document Foundation in future releases of Ubuntu. The Document Foundation’s stewardship of LibreOffice provides Ubuntu developers an effective forum for collaboration around the code that makes Ubuntu an effective solution for the desktop in office environments”

See that and others on the “Supporters List”.

A decent review of anti-trust actions against M$ and M$’s fight against competition around the world is on Ars Technica. For those too young to remember or too busy to be engaged, there is a good outline of events from the legal side. If you have time, please read the documents on US DOJ v M$. They will change the way you think of M$. They did so many things to mess with competition instead of making a good product. There is just no other way to view merging the browser with the OS. It wasn’t an application. They made it a necessary component of the OS and they claimed it could not be removed. That was just the tip of the iceberg. Read how they threatened major OEMs with higher prices or no right to distribute the software.

This was all too little and too late. M$ got a solid monopoly that it wanted and a decade or more of obscene profits rolled in. The “final agreement” can be seen as a rubber-stamp of acceptability on the monopoly even though it was illegally gained. M$ was not punished in any way, just told not to do it again. That might work for an innocent child but M$ is a brutal tyranny, ruled from the top down by evil men anxious to enslave the world. If you consider that M$ rakes in tens of $billions annually and the product is basically copies of documents/CDs/permissions the world really has been working for M$ and not the other way around. They have created something worth a few $billion and have arranged to be paid repeatedly for every PC sold. They even force people who use thin clients to pay multiple times for a licence just for connecting to the PC or server running their software. They do nothing to earn that money. The world has been working for M$.

It is past time to free the world of this evil. Use Free Software. Use GNU/Linux, or FreeBSD, or OpenSolaris, anything but that other OS that does not even have a legitimate name. If the courts refuse to see the light, what is your excuse? Be free. Don’t buy any of M$’s products. Find out what you have been missing. I saw the light 10 years ago. PCs that refused to work reliably danced. Students had good use of IT in the classroom. I became a magician in the classroom. Today I had standing room only in my lab as students thrilled at the performance of all their applications running on one PC that would have laboured to please one user with that other OS (and restricts the number who may even try by the EULA). Be free.

UPDATE At the end of an article linked from the one mentioned above, ARS Technica reported in July 2010 that only 63% of their visitors used that other OS and 6% used GNU/Linux. Clearly, they have different visitors than NetApplications counts. ARS shows 26% use MacOS although Apple says they produce only about 3% of PCs.

I set up an old ’486 PC in the lab to show students how far we’ve come. I explored it myself today for the sake of nostagia. You can check out RedHat 4.2 on DistroWatch (1997):

• no jounalling file-system, just ext2 and FAT
• no window manager, just xdm (way too close to X)
• before GNOME (1999) and KDE (1999) and OpenSSH (1999)
• no bloat either, runs in 16MB

It was a “jarring experience”, this time-warp. /etc I hardly recognized. Most of the scripts are different now. The performance is roughly how a ’486 behaved with that other OS. I have no numbers from that era. The slow processor really hurts X -query terminalserver. You can watch the screen re-fresh…. like pulling down a windowshade. The bottleneck is likely the 10 megabits/s networking. That could redraw 1024×768 in a couple of seconds (16bits) pixel by pixel (video). Boot-time is about a minute. Log-in to a useful desktop on the terminal server is about 10s, so it is faster than what my users were getting from XP on much newer equipment (2002, five years and several steps of Moore’s Law later) as thick clients. OpenOffice.org opens to a usable window in 10s and at 40 words per minute, I see no lag.

This example gives the lie to the idea that we should chuck equipment every few years. This beast is still usable for most purposes at 15 years of age. The graphics is a bit ugly at 16 bits and there are some protocol problems… The “OpenOffice.org 2.4″ logo that flashed on the screen as the programme was starting was flipped top-to-bottom! The next step up, Pentium, say, and 100 mbits/s networking would be absolutely better than what people are experiencing on newer equipment with that other OS. I don’t care about eye-candy. I mean the data-processing people need to get done on a server or cluster somewhere. An older machine can show the pix and receive the clicks indefinitely. If you can buy RAM and other parts for it, why not? Most parts are $50 or less these days and take less than 15 minutes to swap (perhaps a motherboard would be longer). The systems are worth keeping and they do useful work with GNU/Linux.

Of course, newer equipment and newer software is fun but it is not essential to the task of creating, finding, storing and presenting information. Schools can work with decent used PCs quite well.

oldman provoked me when he mentioned the wonderful tools he has to manage PCs/servers in his system. I thought I would describe my tools. Mine cost $0 and work for me so I see high performance/$.
Infrastructure
My network is not unusual. I use a router just as a firewall and router. I put DHCP on my own server so I can control what each machine looks like on the network. I record the MAC for each machine in a database and placed all the relevant statements in /etc/dhcp3/dhcpd.conf. That way every machine gets the same IP address on each boot and the dhcp server updates the DNS server locally so I can address machines by name. There are GUI tools for doing this but my system is so small, I just edit the lists. I use scripts to create the configuration files. All I needed was a list of machine name and MAC and IP address. I took that from the router when I took away its DHCP function and added machines as they came on line. While DHCP/DNS is a management tool system there is nothing remarkable about that. It is fairly normal on any system.
Custom tools
When I want to tweak anything from my chair, I use tools I built from lower-level tools in the GNU system. I use an account on a particular server to control everything in the system. The OpenSSL public key /root/.ssh/id_rsa.pub for that account is stored in /root/.ssh/authorized_keys in every GNU/Linux disc image so that when I restore the image to install GNU/Linux on a machine, I have passwordless logins remotely for that machine. The /root/.ssh/known_hosts file and the host keys prevent man-in-the-middle attacks once the key signature is approved by me. I can examine that by ssh-keygen -l . This gives me reasonable security for a system like mine. I could improve security by adding more layers but this is enough considering the value of our data etc. Most of our data comes from the web and we just keep a local copy. Most really important documents end up on paper and stored in steel boxes somewhere.

How the tools work is simple. I prepend some stuff to any command and it goes to any and all PCs I control. I can run a private process or a GUI app on any machine I control.

e.g. to check the time on all machines in the system:

./all date

I get a column of date-time data and I can spot any laggards. They should all be synchronized so a discrepancy means a machine has a problem like being off at the time of attempted sync, flaky network, or a misconfiguration. If a machine is off-line, I get an indication of that. All messages are prefixed by the name of the targetted machine:

Here is “all”:
#!/bin/bash
for f in /root/scripts/lab/*;do (echo `basename $f`;ssh `basename $f`.example.com $1 ;echo)&done

The directory /root/scripts/lab/ has touched files with the name of each PC in my lab. I have another command which does similar things for every PC on the LAN. $1 is the command following ./all. I could put all on my PATH to make this easier but I like being in my file structure so I can instantly edit anything so all is in a good place. For the classrooms, I can create directories with the names of the PCs therein. If I forget a room number I can use locate nameofpc to find it.

Here is “wall”:
#!/bin/bash
#wol all PCs in lab
wakeonlan -i 192.168.0.255 `cat /root/scripts/macs/*`

macs is a directory containing files named after the PCs. The content of each file is the MAC (hardware address of the NIC).

Here is “sall”:
#!/bin/bash
#shutdown all PCs in lab
for f in /root/scripts/lab/*;do (echo `basename $f`;ssh `basename $f`.rsl.edu “shutdown -h now&exit;”;echo)&done
~

If I want to run a GUI application on any machine, I can use ssh -Y user@machine application and the X connection is forwarded. This is useful for demonstrating to a user who has walked up to my desk or to use a GUI tool for my system administration. I normally prefer text because it is simpler and methods and results are easily saved.

I also have scripts for waking any or all and shutting any or all down. The most frequently used command to ferry around this way is apt-get update;apt-get upgrade which synchronizes the system with the repositories of which I have a cache of used packages on my server. It all works very nicely. I have used similar tools for that other OS and they are all propietary add-ons and because of the lack of a package manager much more complex. No doubt oldman will tell us the wonders of a provisioning system that can install x and y inside a virtual machine on any machine in the system in minutes but I have no use for such a thing. I have not enough RAM per machine. I have 12 machines capable of that on teachers desks and GNU/Linux as a normal installation does everything they need done and more. My problem is getting teachers to use all the capabilities of this system not adding more to it.

To a stock Debian GNU/Linux installation all I need to get this to work is to apt-get install openssh-server and then plug in the authorized_keys as mentioned above. It takes seconds to do that. My cache of the repositories runs at gigabit/s between servers and 100 megabits/s to any client. That’s many times faster than our Internet connection and makes the packaging system APT a joy to use. My cache has 500gB drives so this system could likely work well for a few thousand machines. RedHat, Novell and IBM have toolsets that they claim work for 10K machines. My school with less than 100 does not need that and it would take longer for my successor to adopt to say nothing of the year or more it would take to get that in the budget.

Oh, in particular, for teaching, a teacher should control the PCs used by the students. With this system I can add stuff to view/control the desktop (x11VNC) or I can watch their list of active processes. I can stop any process, too. pkill -u user, pkill -f somecommand, /etc/init.d/gdm stop, all get their attention instantly. The last logs everyone in a GUI off the terminal server in seconds. Very dramatic way to send them on their way. I can also shut down the lab as described above. This sure beats arguing with a student or phoning Mommy. It helps to have a lab where I can view each monitor directly. If all are supposed to be writing a report and one has a video game running, he/she stands out.

Needless to say, the password of root on the main server is a priceless commodity in my system. I will not sell it cheaply…