LWN.net Weekly Edition for October 20, 2011

Hands on with Plasma Active One

Touchscreen tablets are all the rage these days, though Linux support for them has arguably been late in coming. While Android has made a (closed) release for tablets, free software alternatives have lagged, though both GNOME and KDE are working on the problem. In fact, KDE recently announced its entrant, Plasma Active One, which is the first of three currently planned releases of its touch-based user experience. Giving the new interface a test drive would seem mandatory.

While Plasma Active One can be run in a virtual machine, that is not the recommended configuration. Instead, the installation wiki page suggests several recent tablet devices, like the WeTab or ExoPC, as test systems. I happened to have another of the suggested devices, a Lenovo Ideapad, due to the generosity of Intel at the Dublin MeeGo conference, so I gave Plasma Active a try on that.

The project provides two different live-USB images for testing Plasma Active One, one based on openSUSE 11.4 and the other based on MeeGo 1.2. Both boot directly into the interface, which, after sliding the unlock icon, lands you on a "Introduction" activity that includes a bit of documentation on how to use the interface along with a video that demonstrates many of its features (which can also be seen on the project home page). Both live images worked well, though slowly, which may be expected when running from USB. To try to see Plasma Active One running at full speed, I installed openSUSE 11.4 (as the MeeGo live image seemed unhappy with the Broadcom WiFi on the Ideapad) and followed the wiki instructions to install it. While the speed definitely improved, it was still kind of pokey (in particular, applications loaded very slowly), which may be a reflection of the hardware rather than Plasma Active One itself.

Activities

"Activities" are the central organizing principle of Plasma Active, and the demo systems come with an example activity ("Vacation Planning") as well as a blank "My First Activity" to be filled in by new users. The idea behind activities is to group selected applications, bookmarks, contacts, documents, images, and other objects into a collection that can be switched to easily as the user changes the tasks they are working on. Activities are accessed from the activity "wheel" (or "switcher", seen at right) that gets "pulled" to the left from a tab on the right-hand side of any activity screen. The wheel allows the users to "spin" through their activities to one of interest using a swipe gesture. It's a little hard to describe the action of the wheel, but the video shows it clearly early on. Users can touch any visible activity to switch to it, and activities can be created and deleted from the wheel as well.

The wheel view gives both the name and background image of the activity, which makes it fairly easy to quickly pick the one you are looking for. Switching to an activity brings up its "home screen" (an example activity is shown at left), which consists of various elements that have been connected to that activity. So there may be an "Application" window containing the programs that the user deemed important to that activity, a "Bookmark" window with relevant browser bookmarks, and so on. Various widgets can also be attached to an activity for things like notes, clocks, weather, etc.

The elements (windows, widgets) on the activity home screen can be moved around, scrolled, and resized in the obvious ways (obvious to other touch interface users anyway). There are some constraints on the sizes and positions of the elements so that they do not overlap, but the size/position is also limited to a fairly coarse grid. That makes sense for touch-oriented interfaces, but may seem a little too rigid to some. The activity screen itself is not limited in the vertical direction, though, and can be scrolled to access further real estate as needed.

Applications

Applications can be run either from an activity's list or from the launch area. Pulling down the "Peek and Launch" bar (which lives at the top of the screen and contains status icons for battery, network, and others, along with some control elements) will expose a live view of the currently running applications that can be switched to with a touch. Pulling the bar down further exposes the launch area (seen at right), which displays various applications that can be started. As might be guessed, additional running applications or launchable applications can be accessed by swiping the appropriate area to the right or left.

Applications generally run in full-screen mode, which makes sense for a tablet interface even if it still feels strange to curmudgeons like me who are used to seeing more than one application at once. There are various ways to navigate away from a running application. The first is the running-application-panel mentioned above, which can also be used to quit any running application by touching its close icon. Another method is to touch the activities icon in the far upper right corner, which returns you to the current activity's home screen (and leaves the application running in the background). It is a little difficult to get used to, but is fairly straightforward to use once you do.

There are three other icons in the upper right, corresponding to another central idea in Plasma Active: share, like, and connect (SLC). The idea is that compliant applications will be able to offer ways to share (via social media, email, etc.), like (ratings, favorites, ..), or connect (to an activity or calendar event for example) their content. So a user could connect a web site to one of their activities, share a photo with one of their friends (or many of them via a web service), rate a particular document for use in desktop searches, etc. Currently, though, few applications have added SLC functionality.

Many applications work quite well under Plasma Active One, including Firefox, LibreOffice Math, Marble, and others. While those applications are not targeted at touchscreen interfaces, it is still fairly easy to use them (though typing can be painful, see below). The Kontact Touch applications on the other hand have been written with touchscreens in mind. Things like the calendar and contacts applications were very easy to use and seemed to have all of the functionality that one is used to from, say, the equivalent Android apps.

Some glitches

A touchscreen keyboard is available for text entry, though it suffers from many of the same annoyances that other implementations do. It sometimes pops up in places that cover the text entry field—as it did for searching applications in the screen shot at left—though it can be shifted to the top or bottom of the screen using the arrow icon. It is not really suitable for anything other than entering short text. That's a limitation of the tablet form factor, but the Ideapad does have a hardware keyboard which makes things a bit easier. The on-screen keyboard does show each key value as it is pressed, but the reaction time seemed slow. It also doesn't always pop up or disappear at the right times. For someone whose normal interface to a computer is largely through the keyboard, it was irritating, but that may be partially due to the curmudgeon effect.

There were a number of other glitches that I observed with the three different versions (the two live-USB versions and the installed openSUSE version) of Plasma Active One that I tried. Some may be attributable to the hardware (CPU, graphics, touchscreen) of the Ideapad, but likely others are bugs of one sort or another. The image viewer was flaky (at least in the openSUSE versions, the MeeGo version worked fine), the updates of the running application view would sometimes get out of sync when closing applications, sometimes multiple tries were needed to activate an icon or button, the calendar has an irritating habit of popping up when trying to swipe the Peek and Launch bar (the time, which brings up the calendar, is dead center in the bar), and so on. Those kinds of problems are to be expected in a first release, and many are undoubtedly fixed in more recent development versions. Plasma Active is by no means complete—or completely working—but it is an impressive start.

What's most interesting about Plasma Active is that it has taken its own path. Rather than adopt one of the existing touchscreen interfaces, or picking and choosing the "best" parts of Android/iOS/webOS/etc., KDE has come up with its own paradigms for Plasma Active. As more folks start to use it, the interface may change down the road—something that's already been seen between some early demos at the Desktop Summit in Berlin and the current version. Free software is often accused of being imitative, and not innovative, but KDE and the Plasma Active team cannot be pinned with that label here. Whether one likes Plasma Active or not, it is certainly a bold step in an interesting direction.

Comments (8 posted)

Time zone database attacked but finds a new home

Time is relative, and local time doubly so. The average Linux user may not think about time zones other than at installation time or when on the road, both occasions when one has to pick out their current geographic location on a tiny map, so that the OS can adjust the clock to the local standard. But collectively, all of those individual clock-setting actions add up to the need for a global database — one that needs updating far more frequently than many realize. The database referenced by Linux and most other Unix-like OSes came under fire in October, but has now found a more permanent home.

Unix-like OSes internally maintain the system time as "Unix time", which is the number of seconds since midnight January 1, 1970 in Coordinated Universal Time (UTC), but omitting any leap seconds. UTC is essentially Greenwich Mean Time, and does not have the seasonal adjustments for "daylight saving time" or "summer time" used in many regions. Ensuring that all hosts agree on the time is of course critical to the success of many network protocols, but UTC is not convenient for most of the world's users. User-facing time (including filesystem timestamps, cron jobs, the panel clock, and most messaging or calendaring applications) is presented as the system time adjusted to the local clock, based on the current location.

Exactly what the correct adjustment is depends on several factors, though. The boundary lines between time zones are stable in most countries, but they do change periodically, particularly at the provincial or city level. Whether and when a location observes daylight saving time changes more frequently — often based on political and economic factors rather than any objective or scientific reason — and the addition of leap seconds happens only when the International Earth Rotation and Reference Systems Service (IERS) deems it necessary. On top of determining the current time, however, programs often need to look up the local time for past events.

The solution to this multi-faceted problem is the "tz database", which records the difference between Unix time and local time for 405 distinct regions of the globe. Every change of that delta — whether a leap second, a daylight saving time transition, or any other adjustment — is encoded as a rule. The 405 regions represent those contiguous zones where a single rule has governed the offset from UTC since January 1, 1970, although the database contains some historical records that go much further back. For years the tz database was maintained by volunteers Arthur David Olson and Paul Eggert, with the data hosted by Olson and Eggert's employer, the US National Institute of Health (NIH), along with the tz public mailing list to discuss and announce changes to the data.

Linux systems usually provide the database as the tzdata package, which stores entries in /usr/share/zoneinfo/. The zones are broken down hierarchically by geographic region in human readable form, usually named for the country or largest city in the zone, whichever would be less ambiguous. The GNU libc utilities provide a set of command-line programs to work with the tz database: tzselect, zdump, and zic.

Time out

All was well until Olson posted a startling message to the tz list on October 6 (GMT), announcing that a lawsuit had been filed against him involving the tz database — and that he was therefore shutting down the FTP server and mailing list. In the days that followed, it came out that the lawsuit in question had been filed by Astrolabe, Inc, a purveyor of commercial astrology charts, books, and software, and that Eggert was also named as a defendant (although NIH was not).

The lawsuit (which is visible online) accused Olson and Eggert of copyright violation, because Olson and Eggert gathered some of the historical time zone adjustment data from a book called the ACS Atlas. The ACS Atlas was a reference book of historical facts intended to be useful for astrologers creating or working with "birth charts". Its original publisher went bankrupt and Astrolabe purchased the publication rights to the atlas in 2008. The book does not appear to be in print anymore, but the company does sell a Windows-based application called ACS PC Atlas including the same information.

The lawsuit accuses Olson and Eggert of "unlawful reproduction" of data from the ACS Atlas and of having "wrongfully and unlawfully asserted that the information and/or date [sic] taken from the Works is in the 'public domain.'" It seeks a restraining order and a permanent injunction prohibiting both from "unlawfully publishing any and/or some part of the Works" as well as monetary damages. The lawsuit came as a shock to the tz database community, as well as the Unix community in general, in part because the suit claims that including facts from the book in the database deprives the publisher of income — when the references to the ACS Atlas in the comments actually encourage its use.

The fact zone

But the far bigger problem with the suit is that Astrolabe asserts that the facts in the book are protected by copyright. This would seem to be a matter of settled law in the United States, where the courts have repeatedly ruled that facts or information are not copyrightable material. The Copyright Office even has a FAQ entry and a circular [PDF] explaining the distinction, which lists under "What is Not Protected by Copyright"

... although, naturally, actual court decisions are what matter.

The most prominent ruling is Feist v. Rural (1991), in which the US Supreme Court decided that information was non-copyrightable, even when considerable time and effort went into collecting and compiling it in the published work. Rural Telephone Service Company, a telephone cooperative, had sued the phone book publisher Feist for copying 4,000 entries directly out of Rural's own phone book. The court ruled that originality was required for a work to be protected by copyright, but that even in the case of collections of information or facts, only the "creative aspects" of the collection were subject to copyright protection — not the underlying facts themselves.

The 2003 Assessment Technologies v. Wiredata ruling (mentioned in the "Implications" section of the Feist v. Rural Wikipedia page) extended this idea further, saying that is it "fair use" to reverse-engineer a copyrighted work to extract the underlying factual information from it. The 1987 Worth v. Selchow & Righter Company case would also seem to apply; in that case, Worth — the author of a trivia encyclopedia — sued the creators of the Trivial Pursuit board game for bulk-copying his questions and answers, including many typos. The court ruled in Trivial Pursuit's favor, that facts are not subject to copyright protection.

The Astrolabe lawsuit does not claim that the tz database reproduces any of the ACS Atlas verbatim (which it might be argued was copyrightable as "creative expression"), but the underlying data. Tz database fan and blogger Dave Braverman posted an analysis of the lawsuit on October 6, which resulted in a lengthy press-release-like comment from Astrolabe CEO Gary Christen. Christen claimed that the contents of the ACS Atlas were more than "mere compilations of fact" because the authors had used "ingenuity" researching records and reconciling inconsistencies based on their "best judgments and expertise." Braverman replied with a follow-up post dissecting Christen's comments and predicting swift failure in court.

In another wrinkle, developer Curtis Manwaring wrote to the tz list claiming that the lawsuit was in fact just one part of a larger attack aimed at him. Manwaring is the developer of a rival astrology software product, one that uses the tz database in lieu of ACS Atlas, and claimed that the Astrolabe lawsuit is intended to scare him into licensing ACS Atlas data for his own product.

News from the legal front has been quiet for the past week; several members of the community offered to contribute to a legal defense fund for Olson and Eggert, but none has yet been announced. In the meantime, however, the tz database itself quickly found a new home.

ICANN has TZ

On October 14, the Internet Corporation for Assigned Names and Numbers (ICANN) announced [PDF] that it was taking over the hosting and maintenance of the tz database. ICANN cited a request to take over administration of the database from the Internet Engineering Task Force (IETF), and quotes ICANN Chief Operating Officer Akram Atallah as saying "The Time Zone Database provides an essential service on the Internet and keeping it operational falls within ICANN's mission of maintaining a stable and dependable Internet." It does not mention the lawsuit, although ZDNet Australia quoted ICANN's Kim Davies as saying "We are aware of the lawsuit [...] We'll deal with any legal matters as they arise."

Moving the home for the tz database has been in the works for a while, for reasons completely unrelated to the recent lawsuit. Olson announced the need for a new home in August 2009 because he would "be eligible to start drawing a pension in mid-2012". Since that time, discussions have gone on about where to house the database, and ICANN was certainly in the mix, so the move there doesn't come as a complete surprise.

ICANN is best known for managing the top-level domains of the domain name system and IP address allocations, both through its subsidiary entity the Internet Assigned Numbers Authority (IANA), though it also oversees several protocol number assignments, character encodings, and URI standards. Although it works closely with the IETF, the two organizations are not legally linked.

The tz mailing list came back online — from an ICANN server — a week before, on October 7, and numerous list members posted links to mirrors of the original server's FTP contents. Long-time volunteer Robert Elz appears to be the de-facto manager of the database for the time being, but so far it still appears to waiting for an "official" home server from ICANN. Elz told the list that he has been in contact with ICANN, and that the organization has no intention of interfering with the way the project is run. However, Elz also told the list that he is not interested in taking on management of the project long-term, as he — like Olson — is approaching retirement age.

Time++

Based on the available facts, the Astrolabe lawsuit is baffling at first glance. Unless it was a desperation play by a company rapidly going out of business, it would not seem to be worth the effort. Granted, reading the tea leaves of the legal system is tricky at the best of times, but there are multiple court cases directly refuting the plaintiff's central claims — including a US Supreme Court ruling. What hope could there be to collect damages?

That is, unless Manwaring's story is the key piece of the puzzle. If the lawsuit is actually an attempt to damage a competitor by taking out his alternative source for historical time zone data, then the suit makes more sense — particularly the fact that it names the two individuals as defendants, but not NIH. Individual defendants would be more likely to simply take down the database (which they did), in theory denying Manwaring access to it. The suit references a "takedown notice" sent to Olson and Eggert on-or-around May 12 of 2011, well before the lawsuit was filed (on September 30). Manwaring's product, Terran Atlas, released version 2.0 on May 16, and the product page goes into considerable detail about the merits of the tz database over ACS Atlas.

But if shutting down the tz database (rather than collecting damages) was Astrolabe's goal, then it would appear to have backfired. The information is still available (both through FTP mirrors and through updated tzdata packages), and ICANN has far more weight to throw around than most other defendants. Pursuing a similar case against it would have easily-foreseeable results. Just as importantly, if the current lawsuit continues, the community seems poised to come to Olson and Eggert's assistance.

Comments (18 posted)

Fifteen years of KDE

Fifteen years ago, the graphical interface on GNU/Linux consisted mostly of window managers and basic environments like Motif and CDE that came from the proprietary Unix world. As one of the first GNU/Linux-targeted GUI efforts, the development of the K Desktop Environment became a catalyst for rethinking that situation. It's a role that KDE continue to play as the project celebrates its fifteenth birthday, continuing its tradition of innovation and — from time to time — controversy.

Like the Linux kernel, KDE started with an email. "Unix popularity grows thanks to the free variants, mostly Linux. But still a consistent, nice looking free desktop-environment is missing," Matthias Ettrich posted on the Usenet group de.comp.os.linux.misc on October 14, 1996. "There are several nice either free or low-priced applications available, so that Linux/X11 would almost fit everybody needs if we could offer a real GUI."

Ettrich went on to explain that existing solutions were inadequate because of the lack of a common framework and aesthetic. He suggested using Trolltech's Qt framework — then less than year old — and offered a to-do list of needed tasks, concluding:

At first, Ettrich called the project the Kool Desktop Environment, echoing CDE. However, the name quickly settled down to the K Desktop Environment (KDE), with the "K" not standing for anything. The name stood until 2009, when the project rebranded itself as simply KDE, on the grounds that it was no longer just a desktop, but a community of related software projects.

Originally, Ettrich envisioned the first version being released by Christmas 1996. However, the already ambitious project grew in scope, and with the amount of work needed, the first version was not released until 12 July 1998. The news release was concerned mostly with justifying KDE, and included KOffice as part of the announcement.

Gun-metal gray with no anti-aliasing of fonts, like most interfaces of the time, KDE 1.0 was clumsy-looking by modern standards, and not nearly as unified as Ettrich perhaps envisioned — if only because it had to accommodate many non-KDE applications. However, by the standards of the time, it was far ahead of anything else available for Unix-like systems.

The Qt controversy

With its first release out of the way, KDE settled down to a series of minor releases, moving gradually towards the 2.0 release on 23 October 2000, a re-engineering that brought KDE closer to its ideal of a modular, consistent desktop, and introduced Konqueror as its web browser and file manager — a combination that was popular at the time.

However, even before the 1.0 release, KDE was facing criticism because of its choice of the Qt framework. For KDE, using Qt was largely a matter of practicality: project members considered it the best GUI toolkit available, it was free of charge, and it might encourage commercial software developers to work with KDE.

From the start, Qt was available under different terms for commercial and community purposes. Unfortunately, Qt's first community license, the FreeQT License, was considered non-free by the Free Software Foundation and contrary to the Debian Free Software Guidelines, which was then the second major arbiter of software license's freedom. Moreover Trolltech took several efforts to produce a new license that was generally recognized as free.

At any rate, the second license for non-commercial use, the Q Public License, made little difference. It required that all changes to Qt be made available as a patch so that the code base would remain the same for both free and commercial Qt versions. Since this requirement would mix free and proprietary software, the Q Public License was no more acceptable to the larger community than the FreeQT License was.

In response to this dispute, Trolltech and KDE announced the FreeQt Foundation, whose members signed an agreement that a free version of Qt would always be available. Trolltech and its successors were also obliged to produce a new release at least once a year. If it failed to do so, the latest free version would be released under a BSD-style license. The intricacy of the solution failed to satisfy critics, and the GNU project created the Harmony Project (not to be confused with several projects that have used that name, including the recent copyright assignment project founded by Canonical), whose goal was to create a free Qt clone.

Meanwhile, the dispute was continually escalating into a flame war, fanned by the dedicated beliefs of free software advocates and Trolltech's apparent reluctance to move towards a truly free license. In the Debian project, KDE was a main reason for a general resolution to eliminate all non-free software (it failed). The Free Software Foundation responded just as strongly with the creation of the GNOME project to create a desktop that was unequivocally free.

The dispute was abruptly resolved when Trolltech announced in September 2000 that, after consultation with critics, it now considered the GNU General Public License (GPL) a widespread standard, and would use it for Qt's next release. When the third version of the GPL was eventually released, Qt subsequently moved to it without any real controversy. Since the Nokia acquisition of Trolltech, Qt has switched to the GNU Lesser General Public License (LGPL).

But old feuds die hard, and the animosity between GNOME and KDE still occasionally flares to this day, even though the main reason for it no longer applies. The only difference is that, in recent years, arguments usually center on GUI toolkits and design choices instead of licenses.

The "Golden Age" and beyond

No longer concerned with licensing issues, KDE released a series of uneventful minor versions, before making a major break with the past with KDE 3.0 on 3 April 2002. The release featured improved display capabilities and increased customization and internationalization. Even more importantly, it created subsystems of specialized libraries, including ones for printing, personal information, and document and network protocol access that any KDE-compliant application could easily access. More than any previous release, KDE 3.0 fulfilled the original intention of a fully modular environment.

Some developers criticized the massive rewriting of KDE, and some users described the default icons as childish and unbusinesslike. However, for many, KDE 3.0 was the release showed that GNU/Linux was finally near to equaling proprietary desktops. It became a classic desktop, surviving six years with only incremental releases — three times as long as the previous major releases.

However, gradually the KDE 3 series began showing its age, and rewriting became easier than patching. After two years of planning and development, KDE released KDE 4.0 on 11 January 2008. KDE 4.0 went even further than the KDE 3 series in producing modular subsystems for everything from multimedia (Phonon) to personal information (Akonadi) and interfaces (Plasma). It also added a MySQL database for Akonadi's storage of information.

Even more importantly, KDE 4.0 expanded the concept of the desktop by emphasizing virtual workspaces, easily swappable icons and desktop layouts, desktop widgets, hot spots on the screen edges, and an array of other innovations. For the first time, a free desktop could plausibly be argued to have not only caught up with its proprietary rivals, but leaped ahead of them as well.

"Personally, I naively thought that we had really proven ourselves with KDE 3," Aaron Seigo told me five months after the release. "We had taken the promise of KDE 2 and matured it to KDE 3.5.9. And then we were going to attempt to replicate the results of that previous effort and take it to a whole new level."

But that wasn't how the release turned out. Despite all the promise, the release was something of a failure, provoking hundreds of hostile responses. KDE had failed to emphasize sufficiently that KDE 4.0 was a developer's release, and distributions — eager to offer the latest software — ignored the fact. Users were faced with a desktop that lacked many of the customization features of the KDE 3 series and introduced a sometimes bewildering set of innovations. Jokes, such as Seigo's remark that he had just removed desktop icon support in 4.1 (because icons were now part of the Plasma, and no longer confined to a single desktop) only increased the hostility and the confusion.

Three years later, some users still insist that the third release series was the height of KDE development, and that the fourth series continues to lack features found in the third. Still others consider KDE 4.0 as an example of how developers are divorced from the needs of end-users. However, as minor releases have gradually improved the user experience, most users seem to recognize the KDE 4 series as the accomplishment that it is.

Into the future

As KDE celebrates its fifteenth anniversary, KDE 5.0 seems some distance away. Although the KDE 4 series will be four years old in January 2012, it remains flexible enough that its possibilities are not yet exhausted. For example, the abstraction of the interface from the core code into the Plasma subsystem has allowed Plasma Netbook to be offered as an alternative interface on any KDE installation. The just-released Plasma Active One also shows this flexibility as an interface for touchscreen tablets.

In fact, you might say that with the KDE 4.0 series, KDE is at last fulfilling the vision of fifteen years earlier of a modular, easy to use desktop. Adjusting to circumstances and controversies, KDE has weathered four major releases, and looks ready for another fifteen years of equal accomplishment.

[Note: KDE has no shortage of histories online. For example, the KDE History Page lists conferences and a timeline of events and releases, while The Qt Issue and The History of KDE on the UserBase Wiki gives KDE's side of the controversy over the development framework. A screen shot history is available on Saeid Zebardast's blog and from the screen shot page on the main KDE site (which is where the screen shots for this article were obtained). Those interested in KDE 4.0 might look for my "What Went Wrong with the KDE 4 Release?" when the Linux.com archives are restored.]

Comments (78 posted)

Page editor: Jonathan Corbet

Inside this week's LWN.net Weekly Edition

• Security: Convergence; New vulnerabilities in awstats, java, libreoffice, php, ...
• Kernel: Opportunistic suspend; Timer slack controller; Syscall cgroup
• Distributions: Creating a Fedora ARM distribution part 1; OpenMediaVault 0.2, Ubuntu 11.10, ...
• Development: Martus; Cassandra, IcedTea, LibreOffice, STEED, ...
• Announcements: Android 4.0 SDK, ASF on Apache OpenOffice.org, Dennis Ritchie RIP, ...