In Memoriam: the free software projects we lost in 2010

In November I wrote about the apparent demise of the well-known mail delivery agent procmail, which has not been updated since 2001, but is still routinely packaged by Linux distributions. Whatever your feelings about procmail itself, the story prompted a discussion in the comment threads that we revisit periodically as a community: what exactly does it mean for a free or open source project to die anyway?

There is no one "right answer;" the context of the project, its governance, user community, and the opinion of the debaters make for a wide spectrum of definitions. The process of single-vendor project shutting down can instantaneously switch off source code access, online documentation, and mailing lists. More often, the tools and trappings of a project atrophy one at a time; the documentation wiki slips behind the current release, the milestones never make it past "alpha" announcements on the mailing list and onto the Downloads page, and the user forum slowly transforms into a refugee camp where abandoned users help each other patch and shoehorn the aging code into compatibility with newer system libraries.

In most cases, however, the code itself is still available — somewhere. But if we look back at the projects that did close up shop during 2010, it is clear that the source availability factor alone is not always sufficient to regard a project still among the living. In the end, what makes a free software project dead comes down to practical questions. When was the last release? Are there any support plans for businesses? Is there reliable user-to-user support? Is there support for new developers wishing to leverage the code?

For all practical purposes, there are degrees of mortality to be considered. The owner of the code can walk away, shut down the resources, and fire the developers. Projects that meet that fate have little hope unless an entirely new team revives the work from scratch. But almost as serious is when the project owner or leadership pulls all of the developers and puts them on some successor project — however legitimate the ordination is, there is always a risk that the successor project will never see the light of day, and users cannot make the jump until it does.

In any event, it is the end of the year, which puts many of us in a reflective frame of mind, so taking stock of the wreckage from the past twelve months can be illuminating — both with regard to how open source projects die, and the different directions events can take afterward.

The dead and buried

The easiest casualties to identify are those marked by a corporate owner's official press release or a project's clear statement of discontinuation. A few common factors precede many of these cases — generally predictable ones like lack of consumer adoption or legal woes; risks every project endures on a daily basis.

Easily the most high-profile FOSS project to get the axe this year was Google's real-time collaborative editor Wave. Despite a highly-publicized fall 2009 launch, the subsequent releases of much of the code, and worldwide Wave developer events, the search behemoth pulled the plug in August 2010. For reasons that still baffle me (although the "highly-publicized" part is no doubt a key ingredient), this decision was met with sheer joy from many in the technology press, and outright celebration ensued in some darker quarters. The fact that many of those rejoicing continued to describe Wave as an "instant messaging" tool — which it was not — points to poor product management and muddled marketing as critical mistakes on Google's part.

The same blunders may have stricken the Open Source Applications Foundation (OSAF)'s Chandler, a cross-platform email-and-calendaring application, although its end was not nearly as widely observed. The last announced release was made in July of 2009, although commits trickled in until the end of the year, and while the user and developer mailing lists survive, they now consist solely of requests for install-time help. Despite a well-funded benefactor underwriting its development, the project never achieved a fraction of the mindshare enjoyed by Mozilla Thunderbird (and its Lightning calendar add-on), which itself remains a minority player. I'd be willing to wager less than half of the people who read this paragraph knew what Chandler was, much less had tried it.

In contrast, Linux on the PlayStation 3 met a quick demise in 2010, and it was big news. While the firmware that allowed installing Linux wasn't free software itself, of course, it did allow booting the device with a user-installed OS, and was supported by several Linux distributions. It was such big news because PS3's corporate parent, Sony, knifed the project intentionally, publicly, and without remorse. Citing "security concerns," Sony pushed an April firmware update out to PS3s that disabled the "Other OS" feature the devices had supported since their launch in 2006. Rumors were that fear of Blu-Ray piracy enabled through PS3 hacking were the "concern" in question, although no such exploits were ever published.

A peculiar footnote to the PS3 Linux obituary was Sony's sudden announcement of an OpenStep-based application development framework it named SNAP, which was quickly followed by Sony's sudden announcement that SNAP was canceled. The stated idea was to create an open development framework for Apple's iOS, thus prying the tightly-closed lid off of the iPad/iPhone platform to allow in fresh rays of freedom. Considering that the dream of an open homebrew-development community was the initial justification for allowing Other OS on the PS3, SNAP's brief moment in the sun is probably no big surprise.

The LimeWire peer-to-peer (P2P) file sharing tool was scuttled in October, a move dictated by court order. Sources "close to the company" told PCMag that the application will be reborn as a "copyright-friendly service." Because the court order prevents LimeWire from distributing a client capable of uploading or downloading from the Gnutella P2P network, though, there is little chance that the open source "LimeWire Basic" version of the client will return at all.

Not all project terminations were the result of corporate mismanagement or copyright paranoia, however. Linux's HAL hardware abstraction layer, for example, is officially deprecated in favor of udev. Although this transition has been planned since 2008, both that year and 2009 continued to see additional HAL point releases. As of 2010, the major desktop distributions have migrated away from HAL, although several individual applications still pull it in as a dependency. HAL may continue to receive security patches, but its active life is essentially over.

The lost

Speaking of patches and active lives, several large projects fell into the awkward "dead but still claiming lots of users" category, which poses its own unique set of challenges.

OpenSolaris was just one of many Sun projects acquired by the proprietary database vendor Oracle, and although several of the others (Java, OpenOffice, and MySQL) have had their fair share of headaches and battles since the acquisition, OpenSolaris is the only one to be scrapped outright. A leaked Oracle memo announced the move in September, under which upcoming "Solaris 11" releases might be available through a "technology partner program," but the open source version marches straight for the grave.

In November, the Symbian Foundation met an unceremonious end when majority stakeholder Nokia announced that it would re-absorb the Symbian unit and shut down all of the Foundation's web assets. Those assets disappeared on December 17th, though Nokia reportedly still employs the Symbian development team. Officially, Symbian will remain open source software, and what was the Symbian Foundation will morph into a "licensing body" — but the actual source code will disappear entirely sometime in March. One would be excused for thinking that that doesn't sound particularly open source; anyone who needs the code is encouraged to drop an email to contact@symbian.org — a friendly offer, but not one that alleviates fears of abandonment.

The gone but not forgotten

Sometimes, of course, a corporate parent can cut a project loose, and the project can continue to survive or even grow. Such was the case for Etherpad, the web-based collaborative editor acquired by Google in 2009. Google opened up the code right after the acquisition, but snuffed out the service in May of 2010. Prior to the switch-off date, several replacement services sprouted up based on the Etherpad source code — Pirate Pad, PrimaryPad, OpenEtherpad, and more.

In addition to straight derivatives, the existence of forks sometimes makes it hard to determine when to declare a project dead, but at least one project is a plausible candidate in 2010.

The PHP-based content management system (CMS) Mambo suffered an acrimonious leadership battle in 2005 that led to the departure of the bulk of the developers, who started the Joomla CMS. As is often the case is such a fork, the remaining owners of the Mambo trademark and source code copyrights asserted that nothing was wrong and that development would continue unabated. Although that may have been true for a while, here at the end of 2010 it has been a full calendar year since there were any signs of life from Mambo (longer still since there was a release), apart from the occasional Twitter alert that the project's servers had been attacked. Joomla, on the other hand, seems fine.

The great unknowns

The final category is made up of those projects that have disappeared or show no signs of life, but which, for one odd reason or another, are impossible to outright pronounce them as dead.

Take Xandros, for example. The commercial Linux distribution has not made a release since 2007, although it has acquired a handful of other companies since then, which indicates that capital is not the problem. One of those acquisitions was even fellow distribution Linspire, which has also failed to make a release since 2007. It's not clear whether or not the distribution is dead, though the company itself still exists and sells support contracts for existing Xandros Linux users. The company does have other products, but also it went all of 2010 without making a press release. Any new products the company may be developing is being done behind closed doors

Snort also has a corporate parent that continues to do business, but the tool itself still faces uncertainty. Some in security circles worry that the popular intrusion detection system is on life support if not actually terminal. The reason is that the long-discussed 3.0 rewrite, in planning since 2007, still has yet to appear. The project continued to make incremental updates to the existing version of Snort in 2010, but that apparently was not good enough to satisfy the US government, which paid to have a Snort replacement written.

Raindrop was a combined-messaging inbox system developed by Mozilla Labs, and offered an unusual combination of features: merging email, instant messages, and microblogging discussions into a single stream, and intelligently filtering one-to-one, group, and automated messages. Despite optimistic beginnings, the project quietly stopped receiving updates in late spring, and the mailing lists fizzled. A Mozilla Labs developer told me in October that a Raindrop replacement would arrive "soon" ... but it never has.

The XUL-based cross platform media player Songbird did not shut down entirely in 2010, but it did drop all Linux support in April. Shortly thereafter, it looked as though things on the Linux front were going to be A-OK, when a group of contributors announced the Nightingale project that would pick up where Songbird left off. Eight months later, however, and there has still been no code released.

The caveats on these seemingly expired projects vary. One has to give a some leeway to Nightingale; starting a project from the ground floor but with a large, pre-existing codebase is never easy. With regard to Raindrop, Mozilla Labs is explicitly marketed as the experimental wing of the browser maker, where R&D happens, and actual projects come and go. Either project could still awaken from its slumber and lead a long, happy life. Snort 3.0 could drop tomorrow, of course; perhaps Uncle Sam is just impatient, and the 3.0 rewrite is close to perfection. Who knows what Xandros HQ could have up its sleeve; the ISO downloads are cryptically marked as "out of stock" so maybe it's as simple as a missing hard disk.

2010, we hardly knew ye

Looking back at the list of 2010's dearly departed, you see a snapshot of the open source ecosystem as a whole. Some projects, like Google Wave, Symbian, and Chandler, never found the user-base their creators were hoping for. Others, like LimeWire and PS3 Linux, were forced to walk the plank thanks to legal threats from the code-meets-commercial-media arena. Songbird and Xandros were both popular when they were available, but appear to have simply lost support among the people who write the paychecks at their respective companies (and who knows what happened to SNAP, but at the very least we can agree that "too much support among management" was not among its problems). Mambo got taken out by infighting between its leadership and core developers. If you looked at the active open source projects making the news today, you'd likely find the same kinds of problems.

What is interesting to note about 2010's obituary is that there was only one Oracle acquisition among the fallen. Despite the database company's dominance of the news cycle for lawsuits and anti-community practices, it did not actually succeed in killing that many open source projects. Whether that tells you something about the hype factor of the acquisition or the resilience of the free software community is anybody's guess.

The developers behind OpenIndiana, the community-driven replacement for OpenSolaris, would probably say the latter. That brings us to the other potential lesson from 2010: the number of open source projects that survived, in one form or another. Etherpad is positively flourishing, Joomla is more popular than ever now, MeeGo is growing and even expanding into new areas, and even the much-maligned Wave has been resurrected as an Apache project (presumably to the consternation of some members of the FOSS press).

Three years ago I looked at the projects that perished in 2007 for NewsForge. There were nine projects on that year's Big Sleep list, and although this is not an exact parallel (the 2007 article only covered projects I personally had written about during the preceding year), I can't help but notice that only one of them has survived in any form that I can identify today. There is reason to be hopeful about at least three or four of this year's victims.

The difference could be due to random variation, but it is also possible that the community has learned from experience. For example, there have been large-scale dump-the-code-over-the-wall releases in the past that did not work out as well as Etherpad; perhaps Etherpad's continued existence ought to make it a case study for other such "if the community wants it, the community can have it" divestments. It might not even be too late for some of this year's casualties, say, Raindrop and Symbian. Even though both of them have some prospect for survival, good intentions offer no guarantee either will still be here in 2011.