The World Series of Linux: Round 3, The Championship

CMP Channel Test Center conducted its first-ever World Series of Linux, looking at six desktop distributions of the Open Source OS. Over three rounds, they were put through the paces to see if Linux is ready for prime time.

So we get to the final smackdown in CMP Channel Test Center's World Series of Linux. It's Ubuntu vs. Fedora.

Both Linux distributions have very loyal -- and sometimes-fanatic -- user communities, so engineers were very curious about how the two distributions would compare against each other. The field moved back to the Systemax PC, giving Ubuntu the home-field advantage. However, Linux distributions in general are fairly limited in what they can do without a working network connection, so something had to be done to make the system a playable field for Fedora.

In the real world, a solution provider trying to get networking on a PC has two options: compile drivers and hack the OS or install a new card. Engineers picked the easier one, slapping a 3Com Etherlink 10/100 PCI model 3C905C-TXM Ethernet card into a spare PCI slot. As a general rule, 3Com's network cards are very well-supported, and this was no exception.

Both Ubuntu and Fedora use the Gnome desktop environment, making the comparison much easier. Since the two distributions have a similar look and feel, engineers can focus on actual differences without getting distracted by visual fluff.

The final round went through seven innings: installation, networking, creating and managing users, accessing networked Windows shares using Samba, installing third-party software, accessing a handheld device, and connecting to a wireless network. The first two steps just confirmed that the current setup worked for both distributions.

Ubuntu went to bat first. Installation and networking was as easy and efficient as engineers remembered it. That made it quite a relief to be working again with a distribution that makes Linux look easy. Creating users would be familiar to anyone who's created users in the Windows world. User management is found under System Administration. Ubuntu requires a password before opening up the applet. If the logged-in user is part of the sudo group -- group with Administrator privileges -- on the machine, then that password would be accepted. New users can be created or existing users can be modified from the control panel. All users are assigned to a profile: Administrator, Desktop, or Unprivileged. Administrators have full control of the system, and Desktop users have all the control, except for admin functions (like adding new users). Unprivileged users have no access at the start -- administrators can pick and choose through a checklist of options to customize what level of access the user would have.

The list includes options such accessing the CD ROM, floppy, or external drives, administering the system, using the modem, reading system logs, and using sound devices. The level of granularity gives admins some control over what guest users can do on the system. The Desktop profile can also be tweaked to have less privileges, if necessary.

After creating two users, changing around their privileges, and then deleting them, there was nothing else to do for user management. The Windows share in question was a shared folder called Myshare on a Gateway machine running Windows 2003 Server. Engineers mapped the share under Places, Connect, and then entering the machine's hostname (an IP address is also valid input), the name of the share, and a username for the server. After being prompted for a password, there was a shortcut to the mounted share on the desktop and a window automatically opened, displaying the contents. As long as the machine was on the server and the machine information was at hand, this was a snap. The Browse Network button didn't find anything, which was a bit of a disappointment.

The shared folder was fully accessible because that's how the share had been created. Engineers could read and write data to the folder without problems.

For third party software, engineers wanted an application that didn't use the distribution's install mechanism (apt for Debian, rpm/yum for RPM), but an application that downloaded directly from the vendor's web site, and double-clicked to install. Engineers downloaded Geekbench, a benchmarking tool, from Primate Labs. Double-clicking on the application did nothing. Engineers had to open up a terminal window and run the application using the "sudo" command (which tells the kernel to use admin privileges to access this file). A user not in the "sudo" group will be told at this point that this was not something they were authorized to do and the application would not be installed.

This is a bit of a hit-or-miss and depends pretty much on the software vendor, since engineers were able to start the installation for IBM Lotus Symphony Beta by just double clicking on the downloaded file.

Engineers plugged in a Palm V handheld to the PC and used Ubuntu's built-in Palm OS Devices applet to sync data between the device and the computer. Since there are more iPod users than Palm users nowadays, engineers then plugged in an iPod Mini. Ubuntu automatically detected the iPod as an external storage device and placed a shortcut on the desktop. The gtkpod tool is, in essence, an iTunes for Linux, and lets users transfer audio files to their iPod music players. (There are other tools for Zune and other players available.) Gtkpod was easily installed using the Add/Remove Software applet from the main menu bar.

So far, everything was flowing along smoothly and was similar to how Ubuntu's qualification round had been. Up to this point, Ubuntu had hit a grand slam in every inning.

The final test of this round required connecting the PC to a wireless 802.11g network using the Netgear WG111T Wireless USB 2.0 adapter. Engineers plugged the wireless adapter in, half-expecting the wireless to start working and find the network. Everything else had been that easy, so why not?

Sadly, this was the inning that Ubuntu couldn't hit a grand slam. The adapter's blue LED lit up, but there was no sign of it actually seeing a network. Engineers went to the Add/Remove Software applet and installed ndisgtk, a graphical front-end to ndiswrapper, an application that "wraps" Windows drivers so that the kernel can access it. It appeared as "Windows Wireless Drivers" in the installation menu. After installing, Windows Wireless Drivers appeared as an applet under System Administration. Engineers inserted the Netgear CD with the drivers in to the CD drive, opened up the applet, and browsed to the CD to select the .inf file. The driver installed, Ubuntu found the adapter, and after a few seconds, the adapter found three wireless networks, including the one in the test lab. It dropped the wireless connection about a minute later, but engineers were able to get it working continuously without a problem on the second try. While not a home run, Ubuntu still scored some runs.

NEXT: Would Fedora do better and steal back the championship ring?

Installation and networking was also easy and efficient. Thank you, 3com! Since Fedora also uses Gnome, creating the users was an identical process, except for one key difference. Fedora did not ask for the user's password before opening the Users and Group applet; it asked for the root password. Fedora wasn't using a "sudo" group to authorize users. Anyone who knew the root password could access system administration functions, regardless of their current user level. After using sudo in Ubuntu, using root on Fedora seemed to be a potential cause for concern.

Fedora also had the same process for mounting the Windows share. Again, the Browse Network button didn't find anything.

The Geekbench application again did nothing when double-clicked. Engineers had to open a terminal window, login as user, and then run the application. This was another level of access engineers were not altogether comfortable with. Under Ubuntu, engineers typed, "sudo geekbench" and the application ran, while staying within its own user session. Under Fedora, engineers had to login as a root and run Geekbench under root -- not under its own user session. Engineers also had to be logged in as root in order to double-click and install IBM Lotus Symphony Beta.

Fedora also had the same Palm OS Devices applet and had no problem syncing the Palm device. It automounted the iPod and had no trouble transferring audio files through gtkpod.

While Fedora had successfully gone through each inning, engineers' concerns over using the root password had kept it short of matching Ubuntu's success. If Fedora could connect to the wireless network much more smoothly than Ubuntu, it would win the round, hands down.

Engineers went to the Add/Remove Software applet to install ndisgtk. Oddly enough, it couldn't find ndisgtk or ndiswrapper through the applet's search function. Fedora must not be seeing a repository that had the application. After some searching and referencing Web forums, engineers finally found ndiswrapper and managed to install the drivers. While the wireless adapter detected the network on the first try and didn't have any trouble maintaining connection, the effort required to find and install ndiswrapper (and not ndisgtk, the easier application) was just enough to keep Ubuntu ahead.

Even if the sudo/root difference is not as major an issue as engineers felt, the process for wireless drivers had enough stumbling blocks to keep Fedora from decisively snatching away the lead.

A close call, but Ubuntu wins the game and the title of Best Desktop Linux.

Finally, there are a few notes to consider. For users comfortable with sudo, using the root password may seem to be a bigger deal than to users already comfortable with RPM-distributions. This might just be one of those "your mileage may vary" aspects of being a Linux user.

Even though Fedora and Ubuntu both use the Gnome desktop environment, it's not, by any means, identical in look and feel. The main taskbar at the top, for example, is slightly different, with Ubuntu having all its System applets under Administration and Preferences and Fedora having a separate category for power related applets. Even some of the applications have been tweaked.

Engineers used gnome-screenshot to take screen captures of the distributions for each round, and noticed that the application acted differently on Ubuntu and Fedora. It wasn't a case of running with different options -- one or both must have tweaked the code. Ubuntu's tool took a screenshot and then closed automatically. Fedora's tool, on the other hand, stayed open, making it easy to take multiple screenshots without repeatedly going to the taskbar and opening up the tool. It was annoying enough that engineers briefly toyed with the idea of crowning Fedora the winner just for that.

Both communities have a vibrant community that are helpful in posting tweaks and suggestions for how to get something working. While it's not as wonderful as having a dedicated support line that Novell, Xandros and Red Hat would provide, it's much better than randomly searching for a page or two online. The forums are key. If you have a working network connection, either on the Linux PC or a different one, it simplifies troubleshooting exponentially.

From what engineers could tell, there are many ways to install an application on Linux. Ubuntu (and its compatriots) has .deb packages and Fedora and its ilk has .rpm packages, which automatically open up a package installer and runs. There are applications written as a shell script (for example, perl) which simplifies installation. Then there are the installs off the repositories, using the Synaptic Packet Manager or Add/Remove Applications. And finally, even .bin (binary) files can also execute. All these options, while making more applications and tools available, can get a little overwhelming. Just enough to long nostalgically for good old .exe files.