I'm highly doubtful that the courts would rule that jailbreaking is legal given the wording in Bill C-32, especially considering the situation in the US. Although there are no US court cases I'm aware of that decide whether or not jailbreaking is illegal under the DMCA (the United States' implementation of the WIPO Copyright Treaty that Bill C-32 seeks to implement), at least one large company believes that jailbreaking is currently illegal, arguing in the exemption rulemaking proceeding that "The Copyright Office should reject in its entirety the proposed exemption that would permit jailbreaking.". If history is any indication, Apple has the lawyers to make that belief a reality should the question of the legality of jailbreaking ever go to court.

So if Bill C-32 passes in its current form, we are left to rely on its review process, which says that the Copyright Act will be reviewed every 5 years and that the government can make additional exemptions to the TPM provisions. This is similar to the US exemption rulemaking procedure, which occurs every 3 years and adds exemptions to the DMCA in an effort to mitigate its many unintended consequences.

Summary of exemption rulemakings so far

In my opinion, the US exemption rulemaking procedure does not have a stellar track record. To show this, I'll go through some examples:

Watching DVDs with free software

In 2000, during the first rulemaking process, several proposals were made to exempt Linux DVD playing software such as DeCSS from the TPM provisions in the DMCA, such as Tim Lee's proposal. Linux is an example of free software, which many people, including myself, choose to use because it provides us with more freedom than competing products.

The Librarian of Congress responded, denying free software users the right to play DVDs, claiming that "there is no unqualified right to access works on any particular machine or device of the user's choosing" (despite free software being available to access DVDs). The ruling also states that "The reasonable availability of alternate operating systems (dual bootable) or dedicated players for televisions suggests that the problem is one of preference and inconvenience, and leads to the conclusion that an exemption is not warranted.". In effect, the ruling informs free software users that they must purchase Windows or a stand-alone DVD player (in some cases necessitating a TV purchase as well), even though their computer is perfectly capable of playing DVDs.

In DVDs and TPMs: how often is CSS used? I showed that about 98% of DVDs use CSS, which US courts have shown to be a technological measure under US law. And in What C-32 means for DVDs, I showed that libdvdcss (similar to DeCSS) is very likely to be illegal under Bill C-32. So we would have to rely on the exemption procedure to make free software DVD playing legal. And since the US refuses to grant free software users the right to play DVDs, it is likely that the Canadian exemption process would be similarly uncooperative.

Watching foreign or public domain DVDs; skipping DVD ads

In 2003, during the second rulemaking process, the Electronic Frontier Foundation submitted four exemption proposals, which would allow people to:

1. Play copy-protected audio CDs that malfunction to prevent playback
2. View foreign region-coded DVD movies on U.S. players
3. Fast-forward through unskippable commercials prior to movies on DVDs
4. Play and make full use of public domain movies on DVDs

In its 2006 rulemaking decision, the Librarian of Congress rejected all four of the above proposals for various reasons, similar to those listed in the previous section. Twenty other proposals were also rejected, including the Linux DVD playing exemption that was proposed as in the 2000 rulemaking process. Only four proposals were approved.

Having used a Blu-ray Disc containing over 10 minutes of unskippable pre-roll ads, I can see why an exemption for skipping ads is required. Denying such a simple and frustration-relieving exemption does not inspire confidence in the rulemaking process.

"Failing Consumers Completely"

In 2005, because of the many legitimate exemption proposals that the Librarian of Congress rejected, the Electronic Frontier Foundation (EFF) declined to submit proposals in the 2006 rulemaking process and instead published a paper entitled DMCA Triennial Rulemaking: Failing the Digital Consumer. In it, the EFF describes various problems with the rulemaking process, such as these:

Any digital consumer interested in participating meaningfully in the DMCA rulemaking process must first decipher a bewildering array of legal arcana and independently gather considerable evidence. Rather than receiving public comments and engaging in independent fact-finding, as many administrative agencies do, the Copyright Office has instead laid a heavy burden on the shoulders of those seeking DMCA exemptions

The Copyright Office has also established a number of presumptions that discount the legitimate concerns of digital consumers. First, the Copyright Office has said that exemptions will not be granted so long as a work remains available in an unprotected format, even if that unprotected format imposes additional costs and inconvenience on consumers. Second, the Copyright Office has been stubbornly indifferent when DRM restrictions impose increased costs on lawful activities. Third, the Copyright Office has effectively established a general presumption against all lawful consumer activities that do not strike the Office as being sufficiently "important," repeatedly dismissing consumer concerns as "mere inconveniences." Finally, the Copyright Office routinely presumes (on the basis of no independent evidence and contrary to the logic of the free market) that, but for the continued legal inviolability of the DRM technologies that protect them, many forms of digital media would simply be withheld from the market by copyright owners.

The Copyright Office has turned these settled fair use principles on their head in the DMCA rulemaking process. Rather than treating fair use as a forward-looking, evolving regime, the Copyright Office has made it backward-looking, effectively barring courts from addressing the fair use implications of new digital consumer technologies in the 21st century.

The 2006 rulemaking process went on without the EFF's comments and yet again the Librarian of Congress rejected many legitimate proposals, including (you guessed it) playing DVDs on Linux.

July 2010: Still no 2009 rulemaking decision

The Librarian of Congress usually publishes a rulemaking decision in October or November of the rulemaking year. In 2009, the Librarian of Congress instead chose to postpone the rulemaking indefinitely: "the existing regulation...[shall] be amended on an interim basis to strike the reference to the October 27, 2009, termination date for the list of classes of works identified in the regulation.". According to the document, the Register of Copyrights "is in the final stages of making her recommendation to the Librarian of Congress" for the 2009 rulemaking process. As of today (July 5, 2010), there is still no decision (see the rulemaking web site), effectively denying the US public of using any potential new exemptions for over 7 months with no end in sight.

Conclusion and recommendations

The US exemption rulemaking process is plagued with problems that make it virtually useless for its intended purpose. The Canadian government should not rely on a procedure like this to fix problems introduced by Bill C-32 after it has passed. Instead, the government should fix C-32 while it is still under consideration.

Fixing C-32 is not difficult. As I have recommended (along with many others during the consultation process), "it is imperative that any TM anti-circumvention provisions added to Canadian copyright law are directly linked to copyright infringement and that the creation, use, distribution, and import of circumvention tools are not restricted in any way". This means that a given circumvention action would only be illegal if it was done for the purpose of infringing copyright and that the tools required to legitimately circumvent digital locks would be available for those who need to use them, such as free software users. Modifying C-32 to follow these guidelines would avoid the vast majority of unintended consequences that the bill would otherwise have.

I hope that my recommendations will be implemented when Bill C-32 goes to committee. I would be happy to discuss my thoughts with government representatives. I have a meeting pending with my MP, but I would also be pleased to talk to anyone close to the matter. Feel free to use my contact page to start a conversation or arrange a phone call.

Is it legal to jailbreak an iPhone, iPad, or iPod Touch under Bill C-32? By jailbreaking, I mean what is described at http://en.wikipedia.org/wiki/Jailbreaking_for_iOS , not merely SIM unlocking. If jailbreaking is legal, which exception to the anti-circumvention laws makes it legal?

Denver

And the response:

Under the Copyright Act, there is no prohibition against "jailbreaking" an iPad or similar device.

Whether a particular technology, such as one used to prevent the running of non-approved applications on an iPad or similar device, would be captured by the TPM provisions would depend on whether it meets the definition of TPM in the Bill. This would be up to Canadian courts to interpret.

Bill C-32 also includes regulatory powers that enable the government to set out exceptions to the TPM provisions, for instance if it considers that the provisions would unduly restrict competition in an aftermarket sector. In the United States, which has a similar mechanism for introducing exceptions to its TPM prohibitions, an exemption has been proposed in relation to jailbreaking smart phones, like the iPhone.

Erik Waddell
Director of Communications
Office of the Honourable Tony Clement

Background on response's points

By "regulatory powers that enable the government to set out exceptions to the TPM provisions", I believe that Mr. Waddell is referring to the proposed Section 41.21 in Bill C-32, which begins "The Governor in Council may make regulations excluding from the application of section 41.1 any technological protection measure...if the Governor in Council considers that the application of that section to the technological protection measure...would unduly restrict competition in the aftermarket sector in which the technological protection measure is used.".

The United States' "similar mechanism for introducing exceptions to its TPM prohibitions" likely means the process spelled out in 17 U.S.C. 1201(a)(1)(C), which says "during each succeeding 3-year period, the Librarian of Congress...shall make the determination in a rulemaking proceeding...of whether persons who are users of a copyrighted work are...adversely affected by the prohibition under subparagraph (A) ['No person shall circumvent a technological measure'] in their ability to make noninfringing uses...of a particular class of copyrighted works". Bill C-32 also describes a periodic review process in Section 58 of the bill, which will replace Section 92 of the Copyright Act: "at the end of each...period of five years, a committee...is to be designated or established for the purpose of reviewing this Act.".

The exemption that "has been proposed in relation to jailbreaking smart phones" in the United States probably refers to the first of Electronic Frontier Foundation's three proposals in the 2009 rulemaking proceeding to exempt "Computer programs that enable wireless telephone handsets to execute lawfully obtained software applications, where circumvention is accomplished for the sole purpose of enabling interoperability of such applications with computer programs on the telephone handset.".

Additional notes

To be clear, my question was meant to broadly refer to all situations where a hardware and/or software vendor artificially restricts what software can be run on top of its products. I used the iPhone/iPad/iPod Touch as a specific example because it's currently the most visible example of such a situation.

I will be posting another article discussing the US exemption rulemaking procedure shortly. It will further expand on the material in this article to examine how effective such a procedure would be in Canada. Update: I've posted the article: Will exemption rulemaking work for C-32?.

• Pioneer One Episode 1 in Ogg (Theora/Vorbis) - 720p (437 MiB)
• Pioneer One Episode 1 in WebM (VP8/Vorbis) - 720p (265 MiB)

The Theora/Vorbis version will work in Google Chrome/Chromium or Firefox while the WebM version works primarily with pre-release browsers. You may need to use this wrapper page to view the WebM version in-browser.

Transcoding steps

Since not many people are familiar with the process of transcoding (converting from one set of codecs to another), I thought it would be helpful to share the methods I used for converting Pioneer One into Theora/Vorbis and WebM. It's really not as hard as it sounds and it's gotten a lot easier recently with pretty graphical tools.

Theora/Vorbis

While not graphical, ffmpeg2theora is an extremely simple tool that quickly converts video from a wide range of formats into Theora and Vorbis (see the WebM section below for graphical transcoding tools). I chose to use the June 20 nightly ptalarbvorm version (ptalarbvorm is like the new Thusnelda, basically a drastically improved encoder), which is available on the nightly builds page. I ran it from the command line like this:

$ ./ffmpeg2theora-ptalarbvorm.linux -o Pioneer.One.S01E01.720p.theora-VODO.ogv Pioneer.One.S01E01.720p.x264-VODO/Pioneer.One.S01E01.720p.x264-VODO.mkv

The builds page has binaries for Windows, Linux, and OS X. Thanks to j for pointing out the ptalarbvorm build on the theora-dev list.

The whole transcoding process took 2 hours 15 minutes on my Core i5 670, resulting in a 437 MiB file, down from the 1118 MiB H.264 original. The quality of the Theora version looks pretty good so I'd say the ptalarbvorm encoder is quite competitive.

WebM

WebM is a very new video format, launched by Google on May 19 of this year, which I blogged about shortly after its release. Because it is quite new, the WebM transcoding tools are more bleeding-edge and less mature than Theora transcoding tools. However, the tools that are available are easy to use once you've found and installed them.

For virtually any WebM transcoder, you will need the latest version of the GStreamer libraries. If you run Ubuntu, you can add them quite easily using the PPA for GStreamer developers. Once you have the PPA added to your repositories, you should install the libvpx0 and gstreamer0.10-plugins-bad packages, otherwise transcoding tools are likely to complain about missing video/webm and video/x-vpx codecs (thanks to this post for noting the second package).

After the codecs are installed, you'll need a transcoding tool. At first I tried Transmageddon, but the presets turned out sub-par audio. I ended up settling on Arista, a similar project that resulted in better-quality output. To get WebM support, you'll need a newer version than what's available in the Ubuntu repositories. I downloaded the latest version and then installed it manually. You can do this by extracting the archive and running sudo python setup.py install. I chose instead to install locally to avoid modifying my system files by using python setup.py install --home=$HOME/local and then running PYTHONPATH=$HOME/local/lib/python $HOME/local/bin/arista-gtk, but this isn't required. In general, you can run Arista by using the command arista-gtk.

Once Arista is running, just select the Source (the video you want to transcode), the Device (Computer) and Preset (WebM) to get it to transcode to WebM. Then click "Add to Queue" and it will start transcoding, showing you the progress by displaying screenshots of where it is in the video.

Transcoding Pioneer One took about 1 hour and produced a 265 MiB file, a significant reduction from the 1118 MiB of the H.264 original. I did notice some popping in the audio so I think some sacrifices were made to achieve that size, but the video looks pretty similar to the original. Adjusting the encoding parameters would probably correct bugs like that. Over time, I suspect the WebM transcoding tools will become much more mature, allowing users to easily make WebM versions of videos that are indistinguishable from the original.

Arista will also transcode to Theora; just choose Theora instead of WebM in the Preset menu. I chose to use ffmpeg2theora because it allowed me to use a newer encoder; most people will be satisfied with GStreamer's default Theora encoder.

Conclusion

Transcoding videos to free formats like Theora and WebM is a relatively painless procedure these days. There are a variety of transcoding tools that will convert videos without needing to tweak a bunch of settings. Video distributors should spend a few minutes to ensure their videos are available in free formats so that the widest possible audience can view them.

1. Do you know if CSS would be a TPM?

Bill C-32 implements the international standards set out in the 1996 World Intellectual Property Organization (WIPO) Internet treaties, which require protection of "effective technological measures" used by copyright owners to prevent unauthorized use of their work.

Accordingly, whether CSS, or any technology, would be captured by the TPM provisions would depend on whether it meets the definition of TPM in the bill, specifically whether it effectively protects a work. It is worth noting that courts in other countries have already examined this question (including the US, which found that CSS was an effective TPM). It would be up to Canadian courts to interpret whether CSS is a protected TPM in Canada.

2. Do you know if libdvdcss would be illegal under C-32?

Under C-32 it would be illegal to sell or distribute devices that are designed primarily to circumvent a TPM. To determine if libdvdcss falls under this provision, a court would need to determine (i) that CSS is an effective TPM (as discussed in question 1) and (ii) whether libdvdcss is designed primarily to circumvent the CSS TPM.

Erik Waddell
Director of Communications
Office of the Honourable Tony Clement

While the response doesn't clear up the issue definitively, I think it's safe to say that Canadian courts would interpret the TPM provisions as the US courts have. This means that backing up or engaging in fair dealing would be prohibited for 98% of DVDs (see DVDs and TPMs... for how I arrived at this number) under Bill C-32.

I hope that the government will fix Bill C-32 before it is passed by tying the anti-circumvention laws directly to infringement (instead of having a blanket ban with a handful of exceptions like it does now) and removing the distribution restrictions on all circumvention devices as I recommended in my copyright consultation submission. With these changes, Bill C-32 would retain the fair dealing rights Canadians have today for engaging with digital content on DVDs and similarly-encumbered formats, yet it would still provide "adequate legal protection and effective legal remedies against the circumvention of effective technological measures" as required by the WIPO Copyright Treaty that the government wishes to ratify with the bill.

• xentac: "with BillC32 can I buy DVDs and rip them...?"; Tony Clement: "So long as no TPM"
• Drew Wilson: "If you have a home movie recorded on a DVD and you back that movie up..., you’ve broken the anti-circumvention law."; anonymous commenter: "This isn’t correct. Home movies you burn onto a DVD-R/RW are not CSS encrypted, only commercial DVDs are."

By "CSS", the anonymous commenter means Content Scramble System, an optional method of obfuscating the data on DVDs (what some would call DRM). CSS seems to be a "technological protection measure" (TPM) according to C-32 ("any effective technology, device or component that, in the ordinary course of its operation, controls access to a work...") so I will proceed under this assumption. Hopefully someone closer to the bill can comment on the validity of this assumption.

To provide some clarity to the issue of which DVDs are encumbered by CSS (and thus could not be legally backed up or used for fair dealing under C-32), I analyzed 66 DVDs in my household's DVD collection to determine if they used CSS. Here are the results:

DVDs encumbered by CSS

(illegal to backup or use for fair dealing under C-32)

Big Six and mini-majors

"Independent" studios

DVDs not encumbered by CSS

(legal to backup or use for fair dealing under C-32)

"Independent" studios

Summary

• Of the 66 DVDs tested, 57 (86.4%) were encumbered with CSS and 9 (13.6%) were CSS-free
• All of the 45 DVDs from the Big Six and mini-major film studios were encumbered with CSS (they are all illegal to backup or use for fair dealing under C-32)
• Of the 21 DVDs from "independent" studios (those that don't appear to be directly linked to a Big Six or mini-major studio), 12 (57.1%) were encumbered with CSS and 9 (42.9%) were CSS-free

Conclusion

The use of CSS in DVDs is very prevalent, especially among the Big Six and mini-major film studios, which made up over 98.23% of the market in 2009 (adding the top 18 figures from Market Share for Each Distributor in 2009). Given this market share figure, the sample of DVDs I used in testing was disproportionately skewed toward toward "independent" studios, and thus had fewer CSS-encumbered DVDs than the average household's DVD collection would have. I suspect that 98% of the DVDs in an average household's collection would be encumbered by CSS and therefore would be illegal to backup or use for fair dealing under Bill C-32.

I agree with the anonymous poster who said that home movies are not encumbered by CSS and so they can be backed up legally under C-32. Adding CSS to a DVD is not done by default and is difficult to do, requiring a license from the DVD CCA.

Methodology

To determine whether a DVD used CSS or not, I wrote the following C program, which I linked with libdvdcss 1.2.10 and ran after inserting each DVD:

#include <dvdcss.h>
#include <stdio.h>

int main(void)
{
        dvdcss_t dvd;

        dvd = dvdcss_open("/dev/cdrom");
        if (NULL == dvd) {
                printf("cannot open DVD\n");
                return 1;
        }

        printf("scrambled (1=yes, 0=no): %d\n", dvdcss_is_scrambled(dvd));
        dvdcss_close(dvd);

        return 0;
}

I obtained region code information from the debug output of libdvdcss that appeared when I ran the above program.

libdvdcss and Bill C-32

Ironically, downloading libdvdcss would likely become illegal in Canada under C-32 because it states "No person shall...import...any technology...if...the technology, device or component is designed or produced primarily for the purposes of circumventing a technological protection measure" (Section 47 of Bill C-32, proposed Section 41.1(1)(c)(i) of the Copyright Act). This is because the primary purpose of libdvdcss is to circumvent CSS (which I'm assuming is a TPM), even though it can be used for other purposes, such as checking whether a DVD uses CSS or not. I would appreciate clarification on this point from those close to the bill.

It's important to note that the primary use case for libdvdcss is playing DVDs using free software, such as VLC, not to make illegitimate copies of DVDs. Because the DVD CCA will not provide a license for free software players (because they cannot comply with Section 6.2.4 of the CSS Procedural Specifications, which states "All implementations...shall include features clearly designed to effectively frustrate...attempts to discover decrypted confidential CSS Keys"), people who wish to play CSS-encumbered DVDs with free software must necessarily circumvent the CSS.

A major reason for publishing the above list of DVDs which use CSS is that I couldn't find an existing online source that provided similar information. Perhaps it is because of the effective ban on libdvdcss in the US due to the DMCA that such information is not more widely available. It is important for those considering new laws to consider these and other unintended consequences of anti-circumvention clauses. The EFF has compiled a detailed list of such unintended consequences of the DMCA: Unintended Consequences: Twelve Years under the DMCA. Perhaps we can learn from the US' mistakes.

Ubiquity in web browsers

The most critical factor for a video format's success is how many people can view it. Within a year, over 90% of web users will have a WebM-compatible player in their web browser. This is primarily because Adobe is adding VP8 support to Flash. As most users of Adobe Flash Player update regularly and about 93% of web users have Flash, Adobe's VP8 support alone should push WebM usage share close to 90%. Once one factors in all the WebM-capable browsers: Firefox, Opera, and Chromium, that figure will easily pass 90%.

Perhaps most importantly, the ubiquity of WebM decoders in web browsers will allow web developers to deliver their content using a single, royalty-free codec, which has previously been quite complicated. With Flash support for WebM, a developer can simply place a <video> tag around a Flash embed, ensuring that both WebM-supporting browsers and browsers without intrinsic WebM support (such as Internet Explorer and Safari) but with Flash will be able to view the video. By using a single royalty-free codec, web developers can avoid the hassle of encoding their videos in both H.264 and Theora (as is currently required for complete coverage) while at the same time freeing themselves of the legal shackles imposed by H.264.

While not all software players have official timelines for WebM support, it seems very likely that within a year, all of the software I mentioned will support WebM. Relatedly, I hope that Adobe will clarify their stance on WebM support as their public record endorses VP8, but not the Vorbis audio codec, which is required to fully support the WebM format, a combination of the two. Assuming Adobe Flash Player gets full WebM support, Adobe will have finally implemented the solution I proposed to fragmented codec support, for which I applaud them. This would give Flash a second royalty-free audio codec, Vorbis, in addition to the Speex codec it already supports.

Implementations for mobile devices

One of the big reasons Theora hasn't become more widely-adopted is its relative lack of support among mobile devices. No cell phone I'm aware of ships with a Theora decoder by default. Firefox for the N900 renders Theora, but its implementation is not optimized for the hardware, making even 320x240 video difficult to decode in real-time.

WebM is launching with a long list of mobile hardware manufacturers that will support the effort, including ARM, Broadcom, MIPS, Nvidia, Qualcomm, and Texas Instruments. Broadcom plans to launch WebM support in its VideoCore chips in Q3 2010 while Android will support WebM in Q4 2010. This will give WebM a major foothold in the mobile markets, something which no royalty-free video format has had before.

Hurdles

Neither Microsoft nor Apple has indicated they will ship WebM support in their operating systems or mobile phones. On the desktop side, this won't be an issue since Flash's WebM support will make up for Windows' and Mac OS' WebM deficiency. However, on the mobile side, the iPhone's 15% market share could pose a problem if Apple refuses to support WebM. Note that Apple and Microsoft will support WebM for browser video playback (just as they currently support Theora), but only if the appropriate codecs are installed, which is often too much hassle for the average user.

WebM does not appear to specify profiles currently, as H.264 does, which could limit its usefulness in the mobile space where guarantees on decoding complexity are important for rendering smooth video consistently. It seems that WebM is still under development so this situation may be remedied in the future.

Learning from Theora

Theora will remain an important codec for the foreseeable future, as web sites and browsers transition to WebM. While the quality of Theora videos is demonstrably on par with H.264, many leaders of the web choose to believe otherwise, instead claiming Theora takes up too much bandwidth and H.264 is "a vast improvement in quality-per-bit over Theora". With attitudes like these and with no help from Microsoft, Apple, Adobe, or hardware manufacturers, Theora was unable to gain the momentum needed to become a dominant player in web video. Thanks to a large corporate backer that can afford to pay for the necessary hardware and software implementations, a new brand with no bad press or preconceived notions of inferiority, and its royalty-free nature (allowing it to be used in Firefox and Opera), WebM will over time replace H.264 as the new de facto standard for web video.

Never before has such a highly-visible platform denied software developers the right to provide their software to the platform's users. Windows and Mac OS do not prevent users from running the software of their choice, though they may warn the user that the software is not trusted. Ultimately, the choice remains in the user's hands, unlike it is with the App Store, where the choice remains mostly in Apple's hands. With the iPhone, iPod, and iPad, users are not even permitted to obtain software from sources outside the App Store. Running applications of the user's choice is called jailbreaking and is effectively illegal in countries with anti-circumvention laws (like the US DMCA's Section 103) because running such applications requires the user to disable Apple's (weak) restriction mechanisms.

The best way for governments to legislate against the anti-competitive practices of Apple and other monopolistic gatekeepers is to remove the defective legislation that allowed it in the first place. Only when circumvention is permitted, allowing users to install the applications on their choice and not just the ones Apple says they can, will true competition be possible. I've encouraged Canada to do so and I hope other governments will follow suit.

Adobe claims that we are a closed system, and that Flash is open, but in fact the opposite is true.

So according to Jobs, Apple is open and Flash is closed. But according to a WSJ interview with Adobe CEO Shantanu Narayen, Flash is the open one:

I find it amusing, honestly. Flash is an open specification.

So who is really open?

It all depends on your definition.

Adobe's definition of open

Adobe employees tend to define a platform as open if it allows any application or content to be used on the platform (ie. it is open to anyone). By this definition, Adobe's Flash Player is open because it allows you to run any Flash content; no approval is required by Adobe before running Flash content. Conversely, Apple's iPhone OS is not open because it does not allow you to run any application you want, only those that Apple approves.

Apple's definition of open

Steve Jobs defines open in the context of open standards. A platform is considered open if it uses open standards to interface with the outside world, specifically the web. By this definition, Apple is open since it uses and is a primary driver of HTML5, CSS, and JavaScript, the key web standards. And conversely, Adobe is not open because it does not promote open standards, instead promoting the use of the non-standard Flash rather than HTML5 or other existing standards.

Caveats

While Adobe's CEO is correct that a gratis specification of most of the Flash format is available online, this does not make it an open standard as I've previously pointed out. Despite having a gratis specification, there is only a single implementation of Adobe's Flash that will run all Flash content: Adobe Flash Player. Adobe actively encourages developers to only test with Adobe Flash Player (ie. their CS5 site: "Publish content virtually anywhere, using ... Adobe Flash Player for browser-based experiences.") and developers tend to do exactly that (according to a GitHub developer, "One of the biggest benefits of Flash is that you only have to target one implementation: Adobe's.").

As a result, few Flash blobs work in a non-Adobe Flash player, just as few complex Windows apps work correctly in Wine because most Windows app writers only test on Windows. So an "open" specification of Flash is effectively useless because Adobe strongly discourages its use by promoting Flash as a way to avoid the need to test on multiple implementations. Update (2010-05-03): In fact, a non-Adobe player that is fully compliant with all Flash content is illegal, as rtmpdump's takedown has shown. Section 103 of the DMCA allows Adobe to sue any organization that implements RTMPE (a poorly implemented encryption scheme for RTMP) without Adobe's permission, which is exactly what they've done. The whole point of an open specification is that anyone can implement it without paying anyone or asking anyone's permission. In Adobe's view, "open" specifications are purely for show and no one should be allowed to implement Flash but Adobe.

Apple does indeed embrace open standards for the web. It does this for practical reasons: standards allow Apple to precisely control the software that runs on their system as long as it adheres to the standard. This leads to better battery life and touch screen user experience. However, Apple tends to support standards only when it does not control a market (as is the case in desktop operating systems and web browsers); Apple's standards support does not extend to markets where Apple has a relative monopoly. iTunes is notorious for its use of DRM that is incompatible with any devices but Apple's (music DRM is gone now, but video DRM remains and both are provided in formats that are heavily patented by Apple). Apple does not freely publish the specifications need to write applications for the iPhone, iPod, or iPad; developers must pay $99/year for the "privilege" of developing, even if it never earns them a dime. Even if it did publish the specifications, Apple still controls access to the App Store, the only legal way (in most developed countries) to add functionality to the device. This makes it impossible to run an application on handheld Apple devices without Apple's permission.

Update (2010-05-03): Both Adobe and Apple oppose open video on the web. Neither company ships open codecs such as Theora and Dirac. Steve Jobs casually mentions that "A patent pool is being assembled to go after Theora ...". And Adobe has not added Dirac or other codecs to Adobe Flash Player as I have suggested to them. As a result, creating video content that will work on either company's devices requires one to navigate the patent maze of H.264 and similar proprietary codecs. Ben Ward has it right: H.264 is not and cannot be part of an open web stack. For more on H.264, including how the ownership of your work is compromised the second you press "record", see Why Our Civilization's Video Art and Culture is Threatened by the MPEG-LA. For more on codecs in general, and the business motives behind proprietary codecs, see The codec dilemma.

Update (2010-04-30): With respect to Adobe, I think Steve Jobs hits the nail on the head with this statement:

Perhaps Adobe should focus more on creating great HTML5 tools for the future, and less on criticizing Apple for leaving the past behind.

There is a very straight-forward way for Adobe to turn from evil in the eyes of open standard proponents to mostly good: Make Adobe's developer tools produce entirely HTML5 code instead of requiring Flash. Most of the interactive features that Adobe's tools create can be easily implemented with HTML, JavaScript, and CSS. Good examples of this include these 25 JavaScript games.

Of course Apple is not exactly "leaving the past behind". The iPhone Developer Program License Agreement is much more restrictive than any license required to develop even Windows programs and has the same properties as the Windows API: there is no formal specification and even if there were, people only test on one implementation anyway (as a result, Windows apps only work on Windows just like iPhone apps will only work on the iPhone/iPod/iPad). I think Apple is growing further and further away from its roots in 1984, as one jailbreaker so eloquently puts it.

Conclusion

Both Apple and Adobe use the term "open" to refer to themselves, but define the term quite differently. This makes it difficult to give the word "open" any definitive meaning. I think it may be best to avoid the term altogether, as Stallman recommends, to avoid confusion.

We should instead start focusing on the underlying reasons for "open", such as interoperability. When done right, technology allows people to share any type of information using any type of device. Information sharing need not be artificially limited by device-specific formats (ie. Apple DRM) or "universal" runtimes that don't run everywhere (ie. Flash; see Why Flash is doomed). Let's build and use technology that maximizes long-term usefulness, not through information silos, but through standards.

Fiber-to-the-home is gradually replacing these other methods of providing consumer Internet connections. However, not all fiber services are created equal. Here are the main areas where fiber deployments differ (from each other and from cable/DSL), some that you would expect and some you might not:

Bandwidth

Fiber connections have the potential to provide far more bandwidth than DSL or cable. When done right, fiber deployments can easily reach 1 Gbps, while DSL and cable are currently limited to approximately 200 Mbps per subscriber. Few ISPs in North America come close to these speeds, but they are important to keep in mind when considering the long-term viability of a particular type of infrastructure.

Access to the point of presence (POP)

A point of presence is a room or small building where cables from the surrounding area congregate. With cable and DSL, the POP is owned by the cable or telephone company and typically no other companies are granted access to the POP. Companies that resell DSL or cable service generally receive access only to the downstream Internet connection, once it has been converted from telephone lines or cable to some other medium, giving the resellers less control over the service than the telco or cable company.

Fiber deployments tend to provide POP access to any company that wishes to provide an Internet connection to subscribers that connect to the POP. This is called "open access". Examples of open access fiber networks include Cleveland's Case Western Reserve project, Amsterdam's Citynet, and Google's yet-to-be-deployed experimental network.

However, not all fiber deployments are open access. For example, America's largest deployment of consumer fiber, Verizon's FiOS, is not open access. Verizon controls the POPs and does not permit other companies to offer services over their fiber-optic cable. From a business perspective this makes sense, but it is far from ideal for consumers as they are stuck with Verizon's fiber connection unless someone else deploys fiber to their house, which is unlikely and inefficient from an economic perspective.

PON vs. PTP

There are two main types of fiber architectures: passive optical networking (PON) and point-to-point (PTP). With PON, the fiber cables from a neighborhood are aggregated into a single fiber cable before continuing on to the POP. With PTP, each fiber cable goes directly to the POP.

While slightly more expensive to deploy, PTP has many advantages over PON. PTP offers faster speeds thanks to dedicated bandwidth for each user. PTP also allows each user to pick their own ISP (assuming an open access POP), while PON effectively forces all users on a particular aggregated cable to use the same ISP.

Unsurprisingly, Verizon picked PON for its FiOS deployment, as it is cheaper and they don't plan to utilize its full bandwidth capacity or provide other ISPs with access to their POPs. Other deployments, such as Amsterdam's Citynet, use PTP because of its consumer benefits. For more details on PTP and PON, see Ars' The Internet of tomorrow article.

Latency

Typically, consumer Internet connections have 5-20 milliseconds more round-trip latency than business and academic Internet connections. As an example, round-trip latency from a Verizon DSL connection to a business (or a Verizon FiOS connection), all in New York City, averages 23 milliseconds, while round-trip latency from that business connection to a local university is under 2 milliseconds. Round-trip latency from a Verizon FiOS connection to that same business (and to the local university) is about 7 milliseconds.

Providing direct access to the fiber-optic cable in the POP would allow ISPs to offer round-trip latency that is very close to the theoretical minimum (about 100 kilometers per millisecond), as they could control all aspects of the connection, without being impeded by the telcos' or cable companies' latency-adding infrastructure.

Currently, the primary consumer application affected by latency is multiplayer online games. However, other applications are emerging that will benefit significantly from low-latency connections. Video conferencing is just one example.

Terms of service

While the technical details of the infrastructure are important, an Internet connection is not very useful if you are contractually obligated to do very little with it. Many ISPs require that customers agree to an adhesion contract that specifies a list of onerous stipulations on the use of their Internet connection. For example, the Verizon Terms of Service state that customers "may not ... use the Service to host any type of server."

It is important that customers not be restricted from making legitimate uses of their Internet connection, such as sharing freely-licensed software and videos or running their own web or e-mail server. Less restrictive ISP contracts will encourage new applications, like distributed search engines and the Freedom Box that Eben Moglen proposed. With contracts that actively discourage running servers, many interesting new applications are killed during conception because the prospective inventor does not want the applications' users to run afoul of their ISPs' terms of service.

Though it's not a component of fiber infrastructure, terms of service contracts are nonetheless important to discuss. Some companies deploying fiber will also act as an ISP, as Google seems poised to do (though it will also be allowing other ISPs to provide service in its POPs), so bringing up the terms of service issue is relevant.

Summary

Though current fiber deployments are improvements over DSL and cable, many of them do not realize the full potential of fiber, by choosing PON over PTP, restricting access to the POPs, or maintaining restrictive terms of service. Projects like Google's fiber deployment can change all of that. By deploying PTP fiber infrastructure with open access POPs, customers will be free to move between ISPs, forcing owners of other infrastructure (telephone lines, cable, or other fiber) to improve their offerings or face losing market share.

Further reading

• What Google could learn from EU ISPs' experience with fiber
• Amsterdam’s Citynet scores a home run for fibre
• Amsterdam Gets Active With FTTH
• Mixed responses to Google's 1Gbps fiber play
• Line sharing best solution for slow, expensive US broadband

Here's the result of my computer component shopping spree:

Each of the above boxes/packages was shipped separately to me, but all arrived by Friday, March 5 (I placed the order the night of Monday, March 1). The bottom box is the case, the top box is mostly empty but for a serial slot plate, and the middle box is the rest. The yellow envelope contains a DVI cable that I ordered on Tuesday after I realized I forgot it in the original order. Thanks to the wonderfully competitive market that is eBay, I got it for about $5 including 2-day shipping.

Case and power supply

Ironically, the most time-consuming part of selecting components was finding a case I liked. The other components had been decided on weeks ago, but I left finding a case to the last minute, not anticipating how difficult it would be.

The primary problem I have with most cases is that they're larger than necessary. Convention says that a tower must have horizontal 5 ¼" drive bays and use a standard ATX power supply unit (PSU), which is 6 inches wide. Add a little bit of extra room for mounting the PSU and you arrive at 7 inches, the average width of a conventional ATX or microATX case. I chose microATX as there are very few Mini-ITX motherboards for the Core i5 and none of them have the features I want (more on that later).

At first, I searched for alternative case configurations with a standard ATX power supply. One of these is the desktop case, which peaked in popularity in the early 90's. Initially I settled on the In Win IW-D500T.300BL, as it had full-sized expansion slots, but kept the height down thanks to the re-oriented 5 ¼" bays. However, I later determined that the power supply was non-standard so I would be unable to swap in a better one, making this case unsuitable. There are also slim desktop cases like this one, but they limit you to low-profile add-on cards, which are a hundred times more difficult to find than full-size add-on cards and I wanted to be able to easily upgrade my system.

Other case options with standard ATX power supplies included "cube" cases, like this one, and media center cases, like this one. Cube cases have the motherboard lie flat and place the PSU above the motherboard, reducing the case width, but increasing the height. Despite being short, I felt these cases had far too much wasted space. Media center cases optimize the other dimension, keeping the case very thin while increasing its width by placing the PSU sideways next to the motherboard. However, these cases tend to be very expensive, are very wide, and often use low-profile slots.

So I started exploring power supply standards. Some digging revealed other PSU form factors that were better-suited to a minimalist case. I considered TFX (3.4" x 2.5") and EPS1U (3.9" x 1.6"), but it is hard to find a desktop case that supports these. The Rosewill R379-M appears to support TFX power supplies and is quite slim at under 4", but uses low-profile slots and is a beefy 16" deep, which is mostly required to keep the drive bays clear of the motherboard.

In my searching, I stumbled across the Evercase ECE0526, which pushes the limits of compactness for a microATX case. By using the space above the expansions slots for the power supply (see the rear view), this case's width and height are limited only by the dimensions of a microATX motherboard. The depth is also impressive, at 12.5". The inside view shows just how tight this layout is. It may even be possible to use the FSP220-60LE 220 watt, 80 PLUS power supply (discussed later), although the depth of that power supply may cause problems. Despite the optimal smallness of this case, I decided against it because there is no space for expansion cards (low-profile or otherwise), it only supports slim optical drives, which can be difficult to find, and I couldn't find a place to buy one with some brief searching.

I ended up settling on SFX (5.0" x 2.5"), which improved the selection of cases and power supplies, especially those with full-size expansion slots and 5 ¼" drive bays, while keeping case width in check. Of all the SFX cases I looked at, the In Win BK644.BN300TBL had the smallest dimensions (at 10.80" x 5.50" x 12.70"). It was even smaller in volume than the Rosewill R379-M I mentioned earlier, at 754 cubic inches versus the Rosewill's 789 cubic inches. The In Win case happened to look pretty good, too, with a smooth black front panel free of logos and uncovered ports/bays (these are hidden behind plastic doors with magnetic fasteners).

You can see that it has a single, vertically-mounted 5 ¼" bay and two 3 ½" bays, one of which is hidden. Both exposed bays have doors to keep them hidden when not in use, in addition to a door for the 4 USB ports and audio jacks on the front. There is a blue LED behind the large silver power button that lights the ring around it and a yellow hard disk activity LED above it and to the right (when the case is upright), which is right next to a recessed reset button. I very much appreciate cases that include a reset button as there will inevitably be a time when the OS crashes (especially when you're experimenting with its kernel) and you will want to quickly return to a working state without holding the power button for 5 seconds.

I swapped out the bundled 300 watt power supply for an 80 PLUS-certified SFX power supply, the FSP300-60GHS-R. In general, I find that bundled power supplies are low-quality; their fans are cheap and can easily fail. Also, I figured that since I'm aiming for low power, I should at least have an efficient PSU. I hope to eventually reproduce the 25 watt system at Tom's Hardware, which recommends an efficient, low-wattage PSU. Unfortunately, I couldn't get a hold of an 80 PLUS SFX power supply less than 300 watts (or a bracket for the EPS1U/Flex ATX power supply used in the Tom's Hardware article); I'm hoping that what I have is minimalist enough. So far my idle power hovers around 45 watts according my Kill A Watt. This might be closer to 25 watts than it seems, since I don't know Tom's Hardware's methodology for determining power consumption figures. Maybe when I have some time I'll try to pare that 45 watt figure down. This is important for me as I plan to use my system as both a general purpose computer and a torrent/SSH server, leaving it on all the time. The only real issue I have with the FSP300-60GHS-R is that its P4 power connector cable is a bit short. As you can see from interior of the case pictured later, it barely reaches to the top-left of the motherboard. But it still makes it, so the issue is minor.

CPU

I've been following the latest CPUs fairly closely recently as my primary computer, a Dell Latitude D800 with a 1.3 GHz Pentium M processor, has been proving less and less capable of doing what I want. While web browsing remains snappy with Chromium, watching HD videos (even 720p) is not an option; the CPU simply can't keep up. The system also limits experimenting with new OSes and doing system-level programming, as a single core CPU is not really suitable for multiple VMs and it takes forever to compile large programs like Wine and the Linux kernel. I figured it was about time for an upgrade, since I purchased my D800 in 2003, when IE's usage share was 95%, when Firefox was still known as Firebird and hadn't hit 1.0 yet, and before there was a 2.6 Linux kernel or a distro called "Ubuntu".

I had been interested in the Core i7 CPUs for a while, as they provided 4 cores (8 threads) of processing power at a reasonable price. Unfortunately for me, Intel did not provide integrated graphics for the i7. That's why I was excited to hear about the latest generation of Core i5 CPUs, which include a 32nm CPU and a 45nm Intel GPU in the same processor package.

Perhaps I should explain my preference for Intel's integrated graphics. The graphics market is dominated by three players: Nvidia, AMD/ATI, and Intel. To achieve full functionality on Linux, Nvidia's and AMD/ATI's cards require proprietary drivers. Even if I were not opposed to proprietary software on principle (that it should not be used because it does not allow the user to modify it and learn from it), I would be opposed to proprietary video drivers because they have stopped me from fixing performance issues in the past. Specifically, I profiled a particular 2D game and discovered that most of the processing time was spent in the Nvidia driver, which I could not fix. Working on the open source driver might have helped, but it naturally lags behind the proprietary driver, lacking features like second monitor support (essential for using a projector), because Nvidia directs all of their developer resources to the proprietary driver, just like AMD/ATI. As a result, I decided not to purchase Nvidia or AMD/ATI graphics chips for my system. Instead, I chose to use Intel's graphic chip (available in integrated form only), for which Intel provides open source drivers. Intel has been making these drivers available since 2006.

Some other bonuses of using the 32nm Core i5 include AES instructions (for more efficient disk encryption and OpenSSH transfers thanks to kernel and OpenSSL patches) and much lower power consumption than Core i7 chips, as Core i7 platforms require discrete graphics (which consume more power than integrated graphics) and because they use the less-efficient 45nm process. Some have managed to get a whole Core i5 system to use less than 25 watts while idling. Additionally, all Nehalem chips (including the Core i5) support SSE4.2, which includes the CRC32 instruction. This speeds up checksum operations such as those in the upcoming Btrfs, which I will be migrating to when it becomes stable. Support for the new instruction is already in the Linux kernel. Current AMD processors lack all of these features (and also cannot be used with integrated Intel graphics), which are my main reasons for not choosing one, despite their more attractive pricing.

Interestingly, the mid-range Core i5 670 currently has the highest stock base clock speed (3.47 GHz) of all the Nehalem processors currently on the market. This is higher than even the soon-to-be-released Core i7-980X Extreme, which is the first 32nm Core i7 processor and sports 6 cores (compared to the i5 670's 2 cores). However, the 8-core Nehalem-EX server chips that will launch this month may have a higher clock. If not, the Core i5 680, scheduled for release in Q2 2010, will certainly take this crown from the i5 670. But for now I can rest assured that my CPU clock is the fastest Nehalem clock out there, however superficial such a designation might be.

Motherboard

Choosing a motherboard was relatively painless. My criteria was primarily based on the range of ports the motherboard provided, rather than its overclocking ability, power consumption, or performance (the latter two vary little between motherboards). Ideally, my motherboard would support HDMI, DVI, DisplayPort, USB 3.0, SATA 6 Gb/s, eSATA, PCI, IDE, and serial. I wanted PCI and IDE because I have a lot of old components around that use them and serial for a serial console in case the network interface goes down and I'm accessing it remotely (a serial console can also show boot loader and kernel messages). Though less widely-deployed than HDMI, DisplayPort is a much more robust interface and is royalty-free so it seems likely to take off; I already use DisplayPort-capable monitors at work. For more information, see DisplayPort versus HDMI. Just as I got FireWire, USB 2.0, and gigabit Ethernet on my D800 7 years ago, I wanted USB 3.0 and SATA 6 Gb/s to future-proof my investment, though SATA 6 Gb/s was less important as few devices are limited by SATA 3 Gb/s' interface speed. I also wanted an eSATA port so I could access my external 1 terabyte hard drive more quickly; accessing it through the USB 2.0 interface limits the performance.

Of all the options, the Gigabyte H57M-USB3 fit my needs the best. About the only thing missing was SATA 6 Gb/s, which I decided I could add later if I needed it. The ASUS P7H57D-V EVO was another option (it included SATA 6 Gb/s), but it was much more expensive and lacked DisplayPort. The H57M-USB3 is a good deal, and includes a lot of voltage and frequency tuning features that I'm sure overclockers will love.

In retrospect, an H55M-USB3 would suffice for most use cases. It is less expensive, omitting one USB header and "fake RAID" on the the default SATA ports ("fake RAID" is still supported on the two SATA ports Gigabyte adds). Given that I have 10 USB ports hooked up with two USB headers to spare, I doubt most people would miss the extra header.

About the only qualm I have with the H57M-USB3 is that it doesn't appear to support DVI output in the boot screens. When booting the system with only a DVI monitor connected, nothing appears on the screen. However, the OS can use the DVI port once the BIOS is finished, which is exactly how I have it setup now. VGA and HDMI output are both supported in the boot screens.

Other components

With my non-existent overclocking needs, I picked the cheapest 2x2GB RAM I could find. I'll pick up an internal Blu-ray burner when they come down in price. Optical storage is useful for archiving, as an EMP can wipe out the data on hard disks and SSDs (I might be slightly paranoid). For now my optical drive is an external DVD writer. I chose an SSD for my boot disk as the prices have come down considerably and I wanted to see what sub-millisecond seek time was like. So far I'm impressed, though I admit that with all the new fastness, I'm not really sure which component is responsible for each speedup I experience.

Ubuntu 10.04 Alpha

Ever since I switched to a Linux distribution on my primary computer in 2003, I've never looked back. I finally feel in control of my computer (rather than controlled by it) and all of the useful tools that are installed by default have become indispensable (perhaps the best one is a real POSIX shell). With an OEM version of Windows starting at over $100 and Apple's unwillingness to sell me Mac OS unless I buy their hardware (let alone all of these OSes' other problems), choosing Ubuntu was a no-brainer.

I've chosen Ubuntu over other distros, primarily because of its frequent release cycle and excellent, pain-free hardware support. I will concede that I haven't tested other distros very extensively, but when I do try them, their defaults leave something to be desired (ie. no syntax highlighting in "view" on Fedora). I like to stay up-to-date with my distro, installing a new version every 6 months, so having defaults that I don't need to mess with is important to me.

Though the final version is well over a month away, I selected Ubuntu 10.04 for my new system because it was the most likely to support my new hardware, particularly the graphics chip. And I was not disappointed. Ubuntu 10.04 Alpha 3 recognized the Clarkdale GPU immediately, using Intel's open source drivers, and provided me with full access to the monitor settings through Ubuntu's default tool (System -> Preferences -> Monitors), rather than some proprietary manufacturer-specific tool like amdcccle or nvidia-settings. I have successfully used a Dell 1704FPT monitor with VGA and DVI, plus a 37" Panasonic TC-L37S1 over HDMI without issue. The GPU and graphics drivers have no problem handling a combined resolution of 3200 x 1080 (TC-L37S1 at 1920 x 1080 next to the 1704FPT at 1280 x 1024).

One of my reasons for getting a new system was to be able to play HD videos. The Core i5 does not disappoint. I was able to watch Big Buck Bunny in Theora at 1920 x 1080 using Totem, which consistently stayed between 30% and 50% CPU usage. I also watched a bit of the 1920 x 1080 Dirac/Vorbis version of Sita Sings the Blues, which played smoothly as well, though it used more CPU (around 150%, presumably across two cores), possibly because of a less-efficient decoder or a lower-bitrate stream.

Linux distributions have come a long with since I installed Gentoo on my D800 in 2003. System-specific guides for installing the operating system like the one I used are by and large unnecessary these days. About the only component I actively selected for its Linux compatibility was the graphics chip (and that was only because I wanted open source drivers); I just assumed that the other components would work with Ubuntu, and they did.

There were some quirks getting Ubuntu 10.04 Alpha 3 installed, but nothing I wouldn't expect from alpha-quality software. When I initially booted the live CD and it prompted me to upgrade the installer, I agreed, but then it crashed at the final stage of the installation wizard. After rebooting to get the old installer version, I found the old installer refused to enter the partitioning step, claiming that I had no space (my drive was unpartitioned). Partitioning the drive myself with fdisk and re-running the installer solved the problem.

After installing and restarting, I was greeted by a blank screen, but could get to GDM's login screen by pressing Ctrl-F7/Ctrl-F8 a few times. While this should definitely be fixed, it wasn't a show-stopper for me and it's the only real issue I've encountered with the installed system so far. Update (2010-03-21): I did an apt-get upgrade today and this problem has disappeared; I suspect it is fixed in Ubuntu 10.04 Beta 1, released two days ago.

Overall I'm quite pleased with the new Ubuntu. The new theme takes some getting used to, but I think it's for the better. Ubuntu 10.04 ships with Firefox 3.6, a welcome addition given 9.10's default of Firefox 3.5 and 9.04's default of Firefox 3.0. I might not even have to install Chromium, as I did on my D800 for speed reasons, since Firefox is super-snappy on the Core i5.

Conclusion

Building a new computer was a surprisingly painless experience for me, despite not having assembled a complete system in about 10 years. The system does everything I expected and is very snappy. I hope this one will last me as long as my D800 did. Here's a picture of the fully-assembled system next to a Dell 1704FPT 17" monitor for size reference:

After a week of use, I still haven't run into any major issues with Ubuntu 10.04 Alpha. There are a couple of UI quirks that I expect will be ironed out before the 10.04 release, but overall is a pleasant operating system experience. I hope that OEMs like Dell will add more systems running Ubuntu to their lineup when the LTS (long-term support) 10.04 is released. The web browser choice screen is a step forward, but it hides the underlying problem: that most users have little perceived choice in operating systems, which is a problem best solved by OEMs taking initiative in providing users with more operating system choices. Perhaps one day we'll have an OS choice screen (main article), something users should have had years ago.

I'm open to comments or questions about anything in this article. If you'd like to know more about a specific component I selected or if you want to see more photos of particular parts of the system, let me know.