LWN.net Weekly Edition for January 8, 2009

The Grumpy Editor's guide to 2009

Author Charles Stross recently lamented that times have gotten sufficiently interesting that the writing of near-future science fiction is currently impossible. Too much is changing, in too many interesting and unpredictable ways, for anybody to make projections of the future that don't look ridiculous long before that future arrives. Your editor can certainly understand that concern. But, then, your editor's predictions have always looked ridiculous in short order. So there's no reason not to continue with business as usual. Here's a set of wild guesses as to what we might see this year. Woe unto anybody who takes any of this seriously.

Commercial

The net is full of guesses about what the currently-unfolding financial crisis will mean for free software; many of those are wildly optimistic. Your editor is a bit more guarded: the free software community will emerge from this mess stronger than ever, but it may well be a difficult ride. Much of the commercial Linux industry draws a fair amount of its income from the financial industry, and many players in that industry - there should still be one or two left - are likely to be looking to cut their expenses somewhat. So money for little things like critical infrastructure may be a little short until the bonus pool can be brought back to a satisfactory level. Other parts of the economy will be suffering similar pain. All told, economic trouble will make life harder for a number of free software companies - and the people they employ.

Still, the lower cost of free software, along with its flexibility, can only serve to make it more appealing to companies which are trying to develop a successful business strategy for difficult times. The commercial ecosystem around free software should continue to grow, but it may go through some interesting changes along the way.

One thing that will help is that open embedded systems will grow in appeal and become more successful. The iPhone showed what can be done with an interesting hardware platform; at the same time, it has spawned a steady industry devoted to opening up the device. Android-based platforms are quickly showing that it's possible to make an equally (or almost equally) nice device without locking it down in the same way. Awareness of the value of open gadgets will grow, and there will be more of them on the market by the end of 2009. These gadgets may not be as completely open as many LWN readers would like, but they will be a big step in the right direction.

As that happens, your editor thinks that Android will grow in popularity, perhaps to the point where it eclipses other Linux-based handset platforms. Android has no shortage of flaws, but it is a sufficiently well thought-out and developed system that it should be able to attract a real development community - especially if Google opens up its processes sufficiently. And if Google maintains an overly-firm hand on Android, we may well see forked versions aimed at the hardware devices which can run them.

Legal

The pace of GPL enforcement actions will drop as a result of two independent developments: more companies will figure out that free software licensing matters, and developers will decide that they do not want to be part of a high-profile lawsuit. That said, there will be some significant actions on this front in 2009. Meanwhile, the FSF's GPL-infringement lawsuit against Cisco will be settled in a flurry of "win-win" press releases.

GPLv3 migrations will slow, especially among projects that people have actually heard of.

A formerly friendly company may pull an SCO. The sad fact is that failing companies have a certain tendency to look toward their "IP portfolio" as a last-ditch source of revenue. 2009 is likely to see more than the usual number of failing companies; do not be surprised if one of them grasps at this particular straw.

Distributions

Debian Lenny will be released. Now that the ritual firmware flame war and general resolution obligations have been satisfied, it looks like even Debian would be hard put to not get a release out this year. Debian will also make a serious attempt to avoid a repeat of the recent general resolution mess. There will be changes to how resolutions find their way to a vote, and the scripture-like authority of the "foundation documents" may be eroded somewhat.

We still won't know about Fedora's "infrastructure issues". But they'll promise to fill us in as soon as they possibly can. In the mean time, Fedora will crank out two more solid releases, one of which will eventually show up (somewhat transformed) as the next RHEL release.

openSUSE will make it easier for outside developers to maintain packages in an attempt to bolster its relevance in the development community.

Development

The 2.6.33 kernel will be released. In other words, the kernel development cycle will continue at its fast pace, and the numbering scheme will not be changed.

The realtime patch set will be mostly merged by the end of the year. It really has to happen this time. What could possibly go wrong?

After many years of effort, 3D graphics will be a solved problem on Linux systems. We will no longer be second to other systems with regard to functionality or performance - at least, if you buy your video hardware from cooperative companies. Some of the code may still be working its way through the distribution system, but the work will be done.

It will be a make-or-break year for Perl. If the Perl developers cannot either bring new life to Perl 5 or turn Perl 6 into something real, this language will, by the end of the year, have moved well down the road to "legacy" status.

By the end of the year, KDE 4 will have stabilized, and most users will have forgotten what all of the flames were about. Meanwhile, the first pieces of GNOME 3 may be out, but they are likely to be received without great enthusiasm.

The distributed version control system debates will continue in full force through the year. A number of major projects will make the jump to a DVCS, and most of them will go to git. But Mercurial and Bzr (at least) will remain strong contenders.

As a result of declining contributions from Sun and frustration felt by outside developers, OpenOffice.org will be forked. The new project is likely to have some initial troubles - OpenOffice.org is a big program - but it will eventually become the focus of a much more dynamic, community-oriented system.

Conclusion

This article would not be complete without a prediction that free software will be stronger than ever at the end of 2009. Some predictions are easy to make; that has been the trend for many years, after all. Still, it is going to be interesting to see what we will be able to accomplish over the next twelve months. As always, it is going to be fun.

Finally, it will be a hard year for Linux-related media; we have already seen a foreshadowing of the situation with Groklaw's shift into maintenance mode and the recent demises of LinuxWorld.com and Linux.com. It is a hard time to be in the media business in general, and the free software realm offers challenges of its own even in the best of times. That said, LWN appears to be holding steady so far, thanks to thousands of readers who feel that this enterprise is worth supporting. So your editor is able to confidently predict that we'll still be here for the traditional mocking of these predictions in December.

Comments (51 posted)

Debates on the future of Compiz

Is Compiz dying? Possibly not, but the consensus among developers of the compositing window manager seems to be that the project is in serious need of reorganization if it is going to survive.

Founded three years ago, Compiz quickly gained recognition as one of the first projects to deliver 3-D graphical effects on the desktop. Probably its best-known effect is the presentation of multiple workspaces on a rotating cube. The current state of the project dates from the merger of Compiz and Beryl, a fork of Compiz, at the end of March 2007.

Since then, development has been divided into two projects: Compiz, which includes the core functionality and basic plugins, and Compiz Fusion, which includes utilities and more plugins. In theory, the two projects were supposed to merge, but in practice, that has never happened. The projects still maintain separate web sites, mailing lists, and bug trackers, despite the fact that most developers of one project also work on the other.

The community appears to lack both organization and direction, with many developers working on their own branches of Compiz in secret rather than face endless discussion about their goals. Still other developers have drifted away from the project. Under these circumstances, the community has not only been unable to manage a 1.0 release, but, 18 months after the last stable release, is still struggling to complete version 0.8.

More recently, the community has been affected by the withdrawal of Compiz project leader David Reveman. Reveman's departure, apparently made without any official announcement, has led to a lack of leadership, since no experienced Compiz developer appears willing to assume the role of community organizer. Just as importantly, Reveman's refusal to respond to emails after his withdrawal has caused practical difficulties for other developers because much of the Compiz code base is undocumented.

The result is that Compiz, once seen as an exciting, leading-edge project is now being openly denigrated in some circles. For instance, one commenter on a recent Compiz video on YouTube wrote:

The situation came to a head in late December when developer Dennis Kasprzyk announced the creation of a new compiz++ code branch. This new branch is written in C++ as opposed to the C programming language of the main branch, and would require numerous changes in the behavior of plugins. A few days later, Kasprzyk's announcement motivated Kristian Lyngstol, another developer, to begin a thread on the Compiz mailing list on "The Future of Compiz." This thread was echoed in an article called "Compiz is dying and we need to fix it" by Kevin Lange. Since then, discussions about the state of Compiz have emerged on numerous other mailing lists, especially those dedicated to specific distributions.

According to Lyngstol, "there has been the equivalent of no progress since the merger. We've basically been in maintenance mode. The reason for this, from my point of view, is a complete lack of direction and leadership."

Lyngstol sees several reasons for the current state of Compiz. To start with, he suggests that project members have been waiting too long for "something that will change everything," and the result has been too many code branches, many of which are incompatible with each other. "The reality is that all these branches are counter-productive, regardless of how fun or flashy they are," Lyngstol writes. He continues:

Next, Lyngstol notes that the community remains small, with less than 20 people contributing code, if the subscription list for Compiz-Fusion Planet is an indication. In fact, Lyngstol writes, "Unless I'm missing something obvious, we haven't seen a single new core developer that contributes significantly to [the main branch] since the merge. We have, however, lost a few."

Lyngstol goes on to suggest the reasons for the lack of developers. Because the project has no direction, he writes, "all development and design is done as a solo race. There's no way to know whether you can work on something without losing your work because some obscure branch gets merged."

Even worse, the merge of Compiz and Compiz-Fusion that was supposed to happen never has, resulting in a duplication of effort that Lyngstol describes as "messy." Much the same state of chaos exists in the code, which is both "undocumented" and "not particularly pretty." Moreover, when new code is added, its functions "do more than C functions should do." But the basic problem, according to Lyngstol, is that "Compiz is a research project," in a constant state of change and is not focused on producing a stable release.

To solve this situation, Lyngstol suggests a merger of the various code branches — or perhaps, an agreement that some or all are forks — and some serious attention paid to project management. "We should have clear goals for every major release," he writes, "and finding those goals should be the top priority after a stable release. For each point-release in a development series, we should also have a clear goal. This will make it easier to predict releases and for developers to help."

[PULL QUOTE: Perhaps the greatest indicator of the state of the Compiz community is not Lyngstol's critique, but the polite agreement with which it has been greeted so far. END QUOTE] Perhaps the greatest indicator of the state of the Compiz community is not Lyngstol's critique, but the polite agreement with which it has been greeted so far. To date, those who have responded to Lyngstol's posting have quibbled over the details of some of his points while not seriously contesting his overall observations or his suggested solutions.

Another, more unfortunate indicator is that, while posters have agreed that leadership and direction are needed, so far none of them have come forward to offer it. Instead, Lyngstol and several other active developers have gone out of their way to state that, while they would support change, they were unwilling or unable to take on any leadership role.

So far, no one has suggested possible external reasons for the diminishment of Compiz. But it may be that, now that the novelty of 3-D special effects have worn off, few reasons exist to develop them; the few practical effects, such as zooms, are too slight to encourage the majority to move away from standard 2-D desktops.

Another possible factor is that 3-D video drivers that are both stable and released under a free license are taking longer to arrive than anyone anticipated, and their lack reduced users' interest in projects like Compiz that require them.

Still another suggestion was made in an anonymous comment on Lange's article: Perhaps Compiz has served its purpose by proving that the free desktop could surpass Windows or OS X in eye candy. However, not everyone would agree — developer Quinn Storm, for example, posted a comment to the Compiz mailing list in which she makes clear that she thinks that Compiz has that goal, but has yet to reach it.

Whatever the reasons and whatever happens, one consolation is that, in free and open source software, nothing is really lost. But, as things stand now, with no one willing to assume the leadership of the project, a very strong possibility exists that the the Compiz will continue to diminish, with its members aware of the situation but unable or unwilling to change it.

Comments (18 posted)

The Android Dev Phone 1

Your editor's long-suffering spouse will attest that gadgets are never in short supply in the house. Many of them pass below her interest, but a new one has come in which has attracted attention throughout the household: an Android Dev Phone, otherwise known as the fully unlocked version of the G1 phone offered by T-Mobile. This phone is certainly a fun toy, but it has the potential to be a lot more than that.

The details of this device have been well publicized for a while now. It includes a nice touchscreen display, QWERTY keyboard, GPS receiver, accelerometer, 3.2 megapixel camera, and more. The whole thing is powered by Google's Linux-based Android platform. The Dev Phone is essentially the same device as that sold by T-Mobile, but with a crucially important difference: it is unlocked in all senses. This means not just that it can be used with any mobile carrier's SIM, but also that the base operating software has not been locked down. This is a phone for which the entire system can be rebuilt and replaced at will.

The Dev Phone thus joins the OpenMoko Neo Freerunner on the very short list of truly open mobile handsets. This device, though, has the advantage of being a bit more of a finished product with what appears to be a rather stronger software development team behind it. It also, for what it's worth, has some nice hardware capabilities that the Neo lacks: quad-band GSM, 3G (though not on the bands used by your editor's carrier, alas), keyboard, etc. Your editor believes that it will be a successful product.

Over the course of the next few months, your editor plans to dig into this device and report on what he finds. How open is the device really? What does it take to put a new kernel onto it? What might it take to put a different operating system onto it altogether? And, in general, how does this whole Android thing work? Assuming that he does not brick the device early on, your editor hopes to get a real sense for what can be done with this device, how close its software is to what we normally think of as Linux, and where it might go into the future. It should be a fun project.

First, though, one has to get through the stage of simply playing with the new toy. So the rest of this article will be a user-level review of sorts.

The hardware: it feels generally solid. The device is larger and heavier than handsets your editor has used in the past, but that is to be expected. The keyboard works better than one might think given its size; even your relatively fat-fingered editor is able to type with reasonable speed and accuracy. The vibrator lacks strength. The camera seems to take nice photos (for a phone camera), but it is exceedingly slow. As with most color-screen devices, the display is entirely unreadable when the backlight is off. A nice touch with this phone is an indicator LED which blinks when the phone has something to tell you - an unread text message, for example - but the use of the LED seems to be somewhat inconsistent.

Your editor has yet to get a sense for what the battery life would be in the absence of children playing with the device all day long. Complaints about battery life can be found on the net, but it appears that the phone should be able to get through two or three days of moderate usage where the GPS receiver is off most of the time. On the other hand, if you let your kids use it to mess around on video sites, the battery runs down relatively quickly.

On the software side, this phone gets off to a bit of a rough start. It first requires the user to configure the phone to access data service from the carrier, a process which must be done by hand if that carrier is not T-Mobile. Your editor's last new phone recognized the carrier from the SIM and handled this task automatically. More annoying, though, is that the phone requires the creation of a Gmail account as part of its setup process. The fact that one does not have - and does not want - such an account is not relevant. So now your editor has an entry in the Gmail account database which will never be used.

That, of course, ties in to why Google has gotten into this exercise in the first place. There are many features of the Android platform which are designed to tie the user in more closely to services provided by Google. Some features, such as the calendar, are really just an extension of the online offerings. The phone wants to sync the contacts list to...somewhere...and turning the feature off leads to unpleasant behavior. It is possible to use many of the features of the device without connecting back to the Google mother ship, but it's not the natural mode of operation.

Another example is email handling. There is a separate icon for Gmail which just works; that application offers the features (such as threading) provided by that service. One can run a different mail application to connect to a POP or IMAP account somewhere, but it's a separate setup process. Later, with luck, one discovers the improved K9 client, which must be installed separately and which requires one to go through the setup process again. Even with K9, the non-Gmail mail client is not what it should be. There is no threading of messages, many basic commands (refiling messages, for example) are missing, etc. Then there's little problems like refusing to connect to a server if it doesn't think it can trust the SSL certificate and failing to authenticate if the user's password contains special characters. One assumes that this client will improve, or that other clients will be ported to the platform, but, for now, it doesn't seem to be a priority for the Android developers.

More generally, though, the Android software is pretty slick. A fair amount of thought has been given to how interaction should work on this kind of device. Once one gets used to a few specific differences (holding a finger on an item on the screen for a few seconds often brings up otherwise hidden options, for example), navigating through applications comes fairly naturally. Only in some cases do inconsistencies pop up - some applications have different notions for how to zoom in and out than others is one that your editor has noticed. As a whole, the interface comes across as polished and attractive.

That said, use of the display could be improved. On a small display, there will always be a certain tension between getting enough information on-screen and avoiding the creation of headaches through severe eye strain. Different users will do better with small fonts than others. But if Android offers an option to configure default font sizes, your editor cannot find it. So it becomes necessary to manually zoom almost every web page, almost every email, etc. to get a sufficient amount of information onto the screen. That gets a little tiresome after a while.

The "Android Market" offers a wealth of applications, most of which are available as free software or, at least, in a free-beer mode. When browsing applications, one runs into the Android security model, which is oriented around a long set of capabilities which can be granted to applications. A program which needs do things like access the net, obtain location data, change hardware settings, etc. must declare the capabilities it needs; these are then presented to the user at installation time. Most users will probably just say "yes," but it is worth taking a closer look. Your editor decided to decline the installation of a Mahjongg game after being unable to figure out why it was asking for full network access.

Beyond the inevitable games (including one of the worst Tetris implementations seen in a while), there is a wide variety of available applications. The "Locale" tool makes up for the (surprising) lack of the sort of "profile" feature found on almost every handset your editor has ever seen; it performs tricks like using the GPS receiver to automatically change profiles when the phone enters the office or a theater. The "bubble" application (shown on the left) turns the handset into a portable level. There's no shortage of "smart shopper" applications, most of which can read a barcode using the camera and look up prices for items. There is a "power manager" which attempts to configure the device for optimal power use in a number of situations; it provides a basic profile functionality as well, though the user should be prepared to spend some time configuring the options into a workable form. There's plenty of travel-oriented applications which will fetch weather reports, currency rates, or call a taxi.

One notable omission, with both the base phone and the available applications, is voice over IP functionality. This handset should be able to do VOIP beautifully, but almost no such functionality is available. There appears to be a tool for Skype users, but that's about it.

There are a couple of applications that are of particular interest to your editor. ConnectBot is an SSH client which works surprisingly well; the developers are clearly working toward the creation of a tool useful for people logging into Linux-like systems. And the terminal emulator provides that all-important feature: a shell prompt on the device. Even more fun, on the Dev Phone, a simple "su" with no password will yield a root shell.

Playing around on the device, your editor sees that the ARM processor provides a mighty 383 bogomips. It appears to have a little over 100MB of usable memory. It's running a 2.6.25 kernel (known to be heavily modified) with a single loadable module called "wlan." And so on. As useful as the keyboard is, trying to use it to type commands at a shell which lacks a history mechanism gets painful after a while. Time to go looking for an SSH server.

There are other useful applications, of course, such as the one which actually makes phone calls. Like the others, it lacks perfection, but one can only assume that, on a platform driven by free software, that imperfect applications will be improved or replaced. How easy it is to do such things is part of what your editor intends to find out in the coming months. Stay tuned.

Comments (37 posted)

Page editor: Jonathan Corbet

Inside this week's LWN.net Weekly Edition

• Security: Filesystem capabilities in Fedora 10; New vulnerabilities in OpenSSL, samba, xen, xterm,...
• Kernel: 2.6.29 merge window, part 1; The future for grsecurity; Btrfs aims for the mainline
• Distributions: Social networking and the Linux distribution; Fedora Unity F9 respins; FreeBSD 7.1-RELEASE; GNUmed Live CD 0.3.8; Lunar Linux 1.6.4; Tin Hat 20081229; Debian votes to move forward with Lenny release; Debian FOSDEM talks!
• Development: BleachBit: Does GNU/Linux need the equivalent of a Windows registry cleaner?, Firebird Roadmap, AMD releases R600/700 3D code, GNOME DVCS survey, finance apps for Linux, new versions of PyGreSQL, BusyBox, Samba, Hosts3D, Twisted, matplotlib, gEDA/gaf, SQL-Ledger, Lepton, Wine, binutils, RFIDIOt, Valgrind, Mercurial, monotone.
• Press: 2009 predictions, a look at 2008, Palm to unveil Linux-based Nova, screen videos with Linux, NEPOMUK Project review, Splashtop review, Android on the netbook.
• Announcements: vxVistA open-sourced, FSFE New Year's Resolution Fellowship campaign, OLPC changes, Openmoko in 2009, Metasploit eXploits contest, UK Perl tutorials, Python training, audio compilation cfw, FOSDEM GNOME Devroom cfp, LAC cfp, OpenSource World cfp, SeacureIT cfp, uCon cfp, Belgian Perl, FOSS Nigeria, LCA miniconfs.

Next page: Security>>