UpFront

diff -u: What's New in Kernel Development

Zack Brown

Issue #167, March 2008

The 2.4 kernel looks more and more immobile. Except for bug fixes, it no longer seems to be the case that any new code will be accepted. Even clean, well-written, minimally invasive driver ports from the 2.6 tree now are being turned away, as Vitaliy Ivanov recently discovered. He'd ported the adutux driver to 2.4 and submitted it, only to be told by Willy Tarreau that the driver would not be accepted, because no one used the relevant hardware on 2.4 systems. The fact that this may be because the driver has not been available was met with the argument that people who may have needed such a driver probably already have found different hardware to solve their problem. And, Willy added, because the 2.4 tree was not changing so quickly these days, those who did want the patch would have no trouble applying it themselves.

Vitaliy was a bit disappointed and surprised by this. But, in spite of the rejection, Willy and other top hackers still helped Vitaliy get the patch into the best possible shape, in case anyone ever did want to apply it. The patch apparently now will live in Willy's own personal tree, which gathers together 2.4 patches that are unlikely to make it into the official tree.

The hardware4linux.info site has come on-line, providing a database of hardware and its interoperability with the various Linux distributions that exist in the wild. Like similar projects, this one relies on user-contributed data.

It's possible that the Linux-tiny Project will be started up again, under Michael Opdenacker's leadership, but there seems to be considerable opposition. Linux-tiny is a general effort to make the kernel smaller, both in RAM and on disk, and to provide a central location to submit all such patches, so they can be fed to Andrew Morton or Linus Torvalds. But, several folks, including Andrew, felt there was no need for a central location beyond the kernel itself. His feeling is that any patches that can help make the kernel smaller should be submitted to him, rather than to Michael or anybody else.

However, as a lot of these patches already have collected around Michael, he feels he's still needed to help organize them and present them to Andrew or whomever. So, there does seem to be the sense that Linux-tiny is needed, in spite of the fact that folks like Andrew are very much opposed. It seems as though this could shake out either way.

A very interesting new distributed filesystem has hit the scene, created by Sage Weil as part of his PhD studies. It's been under development for a while now, but Sage has just made his first official announcement. As a result, the filesystem is likely to be more stable than other filesystems at the time of their initial announcement; however, because of the lack of testing, users probably should not trust Ceph with their data until it has had a bit more time under the spotlight.

Ceph supports the familiar POSIX filesystem semantics and distributes its data across an arbitrary number of nodes on a network. Data is replicated and rebalanced behind the scenes, so the loss of only a small number of nodes would be unlikely to cause any data loss.

Originally, the filesystem client itself had been done in FUSE, which made for rapid development at the cost of some speed and correctness. One of the reasons Sage chose to make his announcement now is that he has begun work on an in-kernel client, which addresses all the correctness and efficiency issues.

Adrian Bunk wants to take away the Experimental configuration dependency. The idea behind Experimental was that users could choose not to see a large swath of unstable configuration options and, thus, focus only on the options that seemed the most thoroughly tested and reliable. If, during kernel configuration, users clicked on the “Enable experimental features” option, they suddenly would see all the weird stuff that hadn't yet stabilized. The great value of the Experimental option was that it allowed newer code to have the widest possible distribution among users, without putting users in a position to harm themselves by inadvertently enabling a feature that would somehow or other trash their systems.

Unfortunately, according to Adrian, so many necessary drivers still are listed as experimental, that distributions have been enabling experimental features by default in their production kernels. In many cases, these drivers have not really been experimental for a long time, but their developers just never bothered to remove the dependency. So now, users have none of the benefit of being able to turn off experimental features. If they want to use their system at all, in many cases, they are obliged to enable experimental features and hope they don't inadvertently enable something else that is less stable.

It's unclear what ultimately will become of the feature. Clearly, many experimental features in the kernel would have to be removed entirely, if there were no way to hide them from users who wanted only the most solid features. If Adrian does remove the Experimental option and nothing replaces it, all those features may lose out on their current high level of availability to new users. Meanwhile, the drivers that had caused the whole problem by failing to remove their dependency on the Experimental option would get to stay in the kernel, because they are not really experimental.

Distro Share Distribution

Doc Searls

Issue #167, March 2008

Linux Counter (counter.li.org) has been keeping track of many things for many years. One of those things is distro share percentages. Here is how they stacked up, as of December 9, 2007. The data is derived from 147,964 registrations entered and 151,087 values.

Distro Share Percentages

DistributionCountPercent
CentOS1,1900.80%
Debian28,94919.56%
Fedora Core10,4517.06%
Gentoo12,6428.54%
Kubuntu1,8371.24%
Mandrake7,6025.14%
Mandriva2,8701.94%
Red Hat11,3497.67%
SUSE14,7579.97%
Slackware13,1668.90%
Ubuntu19,49013.17%
Others26,78318.10%

LJ Index, March 2008

1. Percentage of users that click on ads at least once per month: 1

2. Number of times per month most of those who click on ads do so: 1

3. Out of three persons who click on advertising, the number most likely to be female: 2

4. Projected billions of dollars (US) in 2008 advertising sales worldwide: 486

5. Projected billions of dollars (US) in 2008 on-line advertising sales worldwide: 44.6

6. Projected 2008 worldwide market share percentage for on-line advertising: 9.4

7. Years that will pass before on-line advertising will overtake radio advertising: 0

8. Years that will pass before on-line advertising will overtake magazine advertising: 2

9. Lines of code searched by Koders.com: 766,893,913

10. Lines of code searched by KruglePublic, in billions: 2.6

11. Code repositories searched by Krugle.com: 600

12. Thousands of projects searched by KruglePublic: 100

13. Number of projects registered at SourceForge: 164,138

14. Number of users registered at SourceForge: 1,744,635

15. Service listings in SourceForge Marketplace at launch: 600

16. Millions of Linux-based mobile handsets sold by Motorola: 9

17. Percentage of Motorola's handset portfolio served by the company's Linux MotoMagx platform: 60

18. Percentage of Linux CPUs running AMD Athlon: 14.71

19. Percentage of Linux CPUs running Pentium 4: 12.15

20. Percentage of Linux CPUs running other Pentiums: 24.97

1–3: AOL

4–8: AdAge

9: Koders.com

10–12: Krugle.com

13, 14: SourceForge.net

15: CNN.com

16, 17: AmericasNetwork.com

18–20: Linux Counter (counter.li.org); numbers gathered by December 9, 2007

The Linux Muse

Doc Searls

Issue #167, March 2008

Convergent Living keeps expanding its portfolio of Companion-branded home electronics controllers, all of which involve “server-less smart appliances running rock-solid Linux”. All are intended to work with the company's own components or with those of many other manufacturers. At the time of this writing, Convergent Living's Integrated Mode Subsystem Drivers supported the following:

  • 21 scene lighting systems.

  • Ten distributed audio/video multiroom preamps (with two “coming”).

  • Five media audio streamers.

  • 14 digital media servers.

  • Three I/serial-based components.

  • Four security panels.

  • Seven automation panels.

  • Five I cameras, plus “almost any streaming MPEG-3 camera”.

It also supported a pile of Ethernet converters; VGA/USB extenders via CAT5; a serial router and communications to thermostats, humidifiers, shade controls, weather stations; and other “environmental” electronics by several manufacturers, over an array of data link types.

Its latest controller is the Companion Muse, which communicates to both the Net and local home electronics over Wi-Fi. It has a built-in Web browser, plus the ability to control home systems either through IP (Internet Protocol) connections or through “translators” that speak through serial, IR and other interfaces.

The Muse weighs just less than two pounds and runs on an 800MHz LX-800 Geode processor. It has an 8.4" TFT Active Matrix 800x600 SVGA LCD touchscreen, talks 802.11b Wi-Fi and plays 16-bit audio through either a built-in speaker or a headphone jack. It's recharged through a desk cradle or USB passthrough.

Configuring and integrating widely disparate home electronics tend to be complex professional work, so Convergent Living sells its components through professional integrators. Meanwhile, as the line continues to expand, it demonstrates the handiness of Linux as a solid platform for integrating just about anything.

Linux as an RTOS

Doc Searls

Issue #167, March 2008

Linux success in the embedded space is well established. In October 2007, VDC reported that Linux held a 40% share among embedded operating system choices by system developers. Smaller share slices were held by commercial OS vendors, in-house, “other” and “no formal OS”. In the Linux wedge, free Linux distributions outpaced paid ones by more than two to one. And, the free side was trending upward, with free distros outpacing paid ones by more than four to one among future embedded project deployment plans.

But, that's just one source of stats. More recently, Embedded Market Forecasters (EMF) came out with a report titled “Embedded Linux Total Cost of Development Analyzed”, which it says is based on interviews with more than 1,300 embedded developers. In its summary, EMF reported the following:

  • “Embedded Linux has achieved design parity with commercial RTOSes for most projects.”

  • “Embedded Linux design outcomes are consistent with the outcomes of projects using OSes from commercial RTOS vendors.”

  • “Use of a commercial embedded Linux OS is more effective than a noncommercial 'in-house' Linux development undertaking.”

  • “Embedded Linux can be used in a mission-critical environment that requires MILS (Multiple Independent Levels of Security) or EAL (Evaluation Assurance Level) certification or POSIX (Portable Operating System Interface) conformance, when used in protected memory under a certified RTOS.”

Dr Jerry Krasner, President of EMF and author of the report, said, “This study shows that designing with an embedded Linux OS can be as dependable as designing with an RTOS.”

They Said It

Data likes to meet, have sex and make babies, just make sure it happens in your hotel room.

—Martin Geddes, psd on Twitter, December 6, 2007

Put it all together, and here's what I see happening. In the next few quarters, low-end Linux-based PCs are going to quickly take over the bottom rung of computing. Then, as businesses continue to get comfortable with SaaS (software as a service) and open-source software, the price benefits will start leading them toward switching to the new Linux/SaaS office model.

You'll see this really kick into gear once Vista Service Pack 1 appears and business customers start seriously looking at what it will cost to migrate to Vista. That Tiffany-level price tag will make all but the most Microsoft-centric businesses start considering the Linux/SAAS alternative.

—Steven J. Vaughan-Nichols, www.desktoplinux.com/news/NS2414535067.html

Sun will be announcing a multi-year award program in support of fostering innovation and advancing open source within our Open Source communities. We'll be providing a substantial prize purse and working with the communities involved to develop the approach that works best.

—Simon Phipps, Sun Microsystems, blogs.sun.com/webmink/entry/getting_paid_to_develop

New Features at LinuxJournal.com

Katherine Druckman

Issue #167, March 2008

If you haven't visited us recently, you may have missed Linux Journal's Gadget Guy, Shawn Powers, and his video product reviews. Each week, Shawn has entertained and informed while giving viewers a peek at some interesting Linux-powered gadgets, such as the popular ASUS Eee PC, the Z2 Zipit Wireless Messenger and the Neuros MPEG-4 recorder. Be sure to come back to see what other cool toys he will get his hands on.

Also, take a look at the section aptly named “Live from the Field” to get some interesting perspectives and perhaps even a behind-the-scenes look at Linux Journal from our very own staff and advisory board. These folks tend to have some useful information to share, and you might even get a look at some of their geek gear. After the holidays, many of us posted photos and videos of our geekiest gifts for all to see. If you haven't seen these, they are worth checking out and can be found at www.linuxjournal.com/microblog.

Drop by and write your thoughts in the comments sections or in the forums. We'd love to hear from you.

What Are They Using?

Doc Searls

Issue #167, March 2008

Each month, we'll be featuring a fun Linux implementation by a notable user. Launching the series is Wendy Selzer. A founder of Openlaw, its open DVD forum and the Digital Effects Clearing House, she also was a star attorney with the Electronic Frontier Foundation, where she led EFF's Digital Television Liberation Front, fighting restrictive government technology mandates with open-source software. These days, she lives near Boston, where she serves as assistant professor at Northeastern University School of Law and fellow with the Berkman Center for Internet and Society at Harvard Law School.

Here's Wendy:

I'm using MythTV to power my home entertainment system. The combination digital video recorder, jukebox, streaming audio server and Web browser is a Debian-based Pentium 4 running MythTV and other free software.

I built this machine when the Broadcast Flag was threatening the continued availability of open high-definition television tuners, but since public interest groups (including the American Library Association, Electronic Frontier Foundation and Public Knowledge) defeated the Broadcast Flag, the hardware is still available, and Moore's Law makes it cheaper all the time. (Full specs at wendy.seltzer.org/mythtv; the large-screen TVs pictured aren't mine.)

The DVR picks up over-the-air television in HD and standard def, recording a mix of programs I've directly selected, TiVo-like “season passes” and those it gleans from searches or community-generated lists. Whenever a “best movie of all time” or nature program comes up, I can time-shift it to fit my schedule. If I'd rather watch the “NewsHour” in half an hour, I can time-squeeze it to fit. Independent video from YouTube and Miro round out the mix.

Ripping my CDs to lossless FLAC files gives me a jukebox from which I can select playlists to listen to on my living-room stereo, stream to the study or office, or move to a pocket. I can record the Metropolitan Opera's Saturday matinee broadcasts (streamripper from a crontab) and pull up Wikipedia pages or libretti alongside.

The system that started as a political statement has become immensely practical (and fun). The general-purpose computer lets me watch media as I want to see or hear it. We just have to make sure the media stays unencumbered and the technologies aren't hampered by ill-designed mandates from Hollywood.