Leading items

How canceled is Firefox OS?

On December 8, TechCrunch reported from Mozilla's annual developer meeting that Mozilla had announced it was putting an end to its Firefox OS smartphone program. The announcement spread quickly—but with quite a bit of variation as to exactly what aspects of Firefox OS were being shut down. A number of tech-industry bloggers, for example, characterized the move as Mozilla "ditching" or even "killing" Firefox OS itself. But the actual statement published at TechCrunch was more limited in scope—although it still leaves many unanswered questions about where Firefox OS is heading next.

The primary complication is that the announcement in question was evidently made on stage at the Mozilla event, and so far no video or transcript of the session has been published. What did make it out of the room at the time were tweets from attendees; TechCrunch quoted Georg Fritzsche's, which said: "Honesty & courage for Firefox OS: back off from commercial shipping, re-iterate as innovation project 'connected devices'." It is not entirely clear from the phrasing of the article whether the TechCrunch story's author was present for the talk.

The story characterizes the announced change as Mozilla saying "it would stop developing and selling Firefox OS smartphones." At a minimum, that could be taken to mean that Mozilla will not be producing any more Firefox OS developer devices. There have been five such devices in the past: two variants of the Geeksphone, two of the the ZTE Open, and one of the Flame. Over the past year, there has been increasing interest in using Google Nexus phones as developer devices. They are easier to acquire, considerably more powerful, and can easily be reflashed to Android if one so chooses. Meanwhile, there have been sixteen consumer Firefox OS phones. Given that level of activity and the de-emphasis on developer devices, die-hard Firefox OS fans could be forgiven for interpreting the announcement in the least-drastic sense: as Mozilla's decision to forgo new developer-device production runs for the foreseeable future.

The TechCrunch story also quotes a statement released by Mozilla's head of "Connected Devices," Ari Jaaksi, who said "we weren’t able to offer the best user experience possible and so we will stop offering Firefox OS smartphones through carrier channels. We’ll share more on our work and new experiments across connected devices soon." Although still a bit on the nebulous side, that comment does seem to indicate that Mozilla is shifting the focus of its Firefox OS efforts, rather than simply not launching yet another model of smartphone.

Some readers may remember Jaaksi from his days at Nokia, where he oversaw the launch of the Maemo platform. While Maemo met an untimely end concurrent with changes in Nokia's upper management, the platform is still remembered fondly by quite a few in the free-software community. Since the announcement, Jaaksi had made several comments on Twitter. First, saying "Mozilla will stop offering Firefox OS phones through carriers. But we will continue improving web experience on smartphones." Later, he added: "FirefoxOS is alive and strong, but the push through carriers is over. We pivot to IoT and user experience."

Citing the Internet of Things (IoT) as the new emphasis for Firefox OS may seem strange, considering how smartphone-driven the project's marketing has been so far (it was initially touted, after all, as the mobile OS that would power inexpensive phones in developing economies). But support for other device types has been included for quite some time; the Mozilla wiki highlights smart TVs in particular, with passing references to tablets and smart watches. There have been occasional glimmers of other product types; in 2014 we reported on thin-client desktops like VIA's APC line.

But these alternative device platforms never seemed to attract much attention, and they were not the small, physical-computing device types most people mean when they use the term IoT. Only this past week has there been a report of an IoT-style device built with Firefox OS in mind: the CHIRIMEN, a single-board computer developed in Japan that includes a set of GPIO and I2C headers. The FirefoxOSCentral blog reported on its existence on December 5, though little in the way of hard information is available yet. The CHIRIMEN appears to be produced independently, but it is possible that the development team (called Mozilla Factory) has some sort of relationship with Mozilla: the About describes it as "Organized by Mozilla Japan". It can also be rather hard to reconcile IoT with Firefox, since so many IoT products are embedded systems without displays or input devices.

For its part, FirefoxOSCentral responded to the December 8 announcement by saying there was "no cause for alarm:"

To be fair, whether any news is "cause for alarm" is subjective and personal. FirefoxOSCentral may be more invested in the OS project than the average blog, but the fact remains that the details in the TechCrunch story are fairly sparse. In addition, they answer few of the pertinent questions, like whether or not any Mozilla developers have been re-tasked to other projects or whether the Firefox OS roadmap has changed. There simply have not been clear statements from Mozilla on these practical matters, and it may be quite some time before any are provided. The developer event runs until December 12, and those involved may decide to spend additional time preparing any further announcements, given how this most recent one played out.

For the time being, Firefox OS automated builds and QA tests are still happening on a regular basis, with reports appearing in the recently launched "Foxfooding" discussion forum. Asa Dotzler has also been regularly posting changelog reports to the forum. While commits are not overwhelming in their volume, they do appear to be steady. Some individuals at Mozilla have also posted reassurances that development continues unabated. On top of that, one cannot help but notice that the schedule for the developer event contains a heavy dose of Firefox OS sessions.

In recent years, Mozilla has developed a bit of a reputation for making surprise announcements that incite some degree of backlash and then subsequently require clarification—see the recent Thunderbird-abandonment story, for example. So it is understandable that the public might take an obscurely worded statement like that in the TechCrunch story and assume the worst. But it also remains possible that Mozilla's plans for Firefox OS have actually changed very little, and it will simply be several more days or weeks before the clarified story is released. The uncertainty can be frustrating for those who care about the project, but that is one of the side effects of a news story built on top of a single, brief source report.

The most that can be said now is that changes will surely be coming down the pipe for Firefox OS, but that no one outside of the decision makers at Mozilla know for sure what those changes will be.

Comments (11 posted)

Wireshark 2.0: Now with Qt

The headline feature for Wireshark 2.0, which was released on November 18, is the switch away from GTK+ and to the Qt framework, but there is more to it than just that. The bulk of the changes to Wireshark—the venerable free-software network packet sniffer that started out as Ethereal in 1998—come under the heading of user-interface improvements, but that leads to some improved functionality as well.

The switch to Qt was announced back in October 2013, so it has taken around two years to complete the switch. The main reason behind the change was to better support Mac OS X, which is something of second-tier platform for GTK+. In addition, Qt will make Android and, possibly, iOS versions possible as well. Based on a blog post from Wireshark founder Gerald Combs, Qt has also helped make the interface function "much more smoothly".

Combs noted that the new interface looks largely the same as the previous version, which is intentional.

At its core, Wireshark provides a way for users to capture packets on the local network, then to display them for analysis. The basic three-pane display, with a pane for a list of packets captured (i.e. the "Packet List" pane) and two that show details of the currently selected packet, looks much the same. The packet-specific panes show a decoded version of the packet that splits out various fields in it (Packet Details), while the other shows a hex and ASCII version of the packet (Packet Bytes). There is also a toolbar, display filter entry box, and a menu at the top (as seen at right for 2.0.0rc2 from Combs's post).

There are some things that have changed in the analysis interface, however. Packets related to the one selected in the packet list now have icons to indicate that status. For example, DNS requests and replies have left and right arrows and TCP packets that have been acknowledged have a check mark next to them. In addition, the packet list scrollbar shows a "minimap" of the color of packets nearby in the list—similar to the minimaps in modern text editors. When combined with rules that display different types of packets in various colors, it can help find more interesting portions of the captured packets. The minimap from the screen shot above can be seen at left.

In a webinar [YouTube] given on November 12, Combs and Laura Chappell demonstrate some of the features in the new interface. Many things have been streamlined in the Qt-based interface, they said. But, the GTK+ interface will still be supported until the next stable release, which will be 2.2—odd minor numbers are for development releases.

Some of the examples shown in the webinar were things like an improved interface to choose a saved filter to apply to a capture. Previously that required bringing up a separate window that listed all of the saved filters to choose from; now that can be done directly from a menu just to the left of the filter entry box in the main window. Hiding and showing columns in the packet list can also be chosen directly from a menu that comes up when right-clicking the column headings. The interface for setting coloring rules has also been improved so that colors can be chosen from a "picker" rather than having to enter color names.

There has also been some cleanup done in the toolbar to remove clutter, as well as rearranging the main menus somewhat to put functions in more natural locations. For the graphs and statistics windows, there is no longer a step to choose settings before seeing the output; those settings are available at the bottom of the window, so if the standard output with no filter is all that's needed, users get that immediately.

There are many more keyboard shortcuts in Wireshark 2. The full list of those shortcuts can be found from the "Help" menu. In addition, individual windows have their own shortcuts, which can be listed from the window itself.

The capture screen, which initiates the packet capture process, has been streamlined as well. There are now small graphs of packet activity for each interface (called sparklines) that can be double-clicked to start a capture on that interface. Those can give users a quick idea of how much traffic is being seen on each interface in the system. Many of the options in the somewhat busy GTK+ capture window have moved to a separate options window (which has its own sparklines).

There is plenty more to the interface that is shown in the hour-long webinar. It is worth a look for those interested in more information. A zip archive of the trace files used in the demonstration are available as well.

Wireshark 2 has been localized for seven languages for the release. Chinese, English, French, German, Italian, Japanese, and Polish versions are available and more translations are currently being worked on. The project uses a Transifex site for that translation work and Combs invited those interested to join in the efforts.

In addition to the upgraded interface, lots of bugs were squashed in the release, including a number of security holes. Over the years, Wireshark has gotten something of a bad reputation for its security track-record. As the Wireshark security page indicates, the program is tasked with dealing with enormous amounts of user-supplied data using "dissectors" to parse and decode many different formats, including experimental ones. That's not an excuse for the large numbers of bugs that continue to be found, but most of the code in Wireshark is in the dissectors, which means that the bulk of the bugs found will be security issues.

There are a number of things that can be done to reduce the damage that a capture (or, worse, an attacker-supplied capture file) can do in the presence of dissector bugs. For example, Wireshark should never be run as root and captures can be done with a simpler program (e.g. tcpdump) and then imported into Wireshark running in a limited user account or dedicated machine (virtual or otherwise). In addition, the page lists several ongoing efforts to find and remove bugs from the code, which are further fleshed out on a dedicated security development page.

Overall, Wireshark 2 looks like a solid release that brings a lot of interface improvements. That should make it easier to work with and allow users to be more productive in their network troubleshooting and analysis.

Comments (15 posted)

Hands on with the pi-top

In the do-it-yourself community that has grown up around the Raspberry Pi, the pi-top has been generating buzz for over a year. As the name implies, it's a laptop that runs using a Raspberry Pi. In addition, it is modular, consists of free hardware and software (except for firmware on the Pi's GPU) and ships in a kit for users to assemble—all of which amply justifies the buzz. Moreover, when the pi-top finally shipped last month, it proved to be not just the promised portable and wireless controller for Maker projects, but a low-end laptop suitable for web browsing, email, and light productivity tools. Although minor improvements are needed, the only major feature that pi-top has yet to deliver is the educational software originally promised.

Pi-top was founded as a for-profit company by Jesse Lozano, a self-taught developer, and Ryan Dunwoody, an Oxford engineering graduate. Both had experimented with the Raspberry Pi and Lozano recalls that, when the two met in 2014:

The company originally intended to start shipping in May 2015. But, like many first-time manufacturers, Lozano and Dunwoody found the process more difficult than they had expected. In the end, the simplest course was for Dunwoody to move to China for five months to oversee the process. By the time pi-top began shipping in early November, the company had several thousand pre-orders to fill.

What they are now shipping is a laptop with a 13.3 inch screen with 1366x768 resolution. A strip above the keyboard slides back to reveal the interior. Inside are cables and two printed circuit boards, a model B+ Raspberry Pi with one gigabyte of RAM and the Hub, which includes the battery and video connections. Both boards screw into place along two tracks on the bottom of the case. In theory, both boards can be replaced as needed, although, considering that the Hub is configured for use with a Raspberry Pi, some hacking might be needed to replace it with another single-board computer. The snap-together case is available in green and black, and its parts can be produced on a 3D printer. Battery life is advertised to be ten hours.

At the same time, pi-top has begun a fundraising campaign for its next product, the pi-topCEED, a pre-assembled $99 desktop computer intended for education. At the time of this writing, the campaign has more than tripled its goal of $50,000 and has over 1800 backers. While these numbers are tiny compared to the shipments of major computer manufacturers, pi-top appears to have gone from a start-up to a modestly successful small business in less than a year.

Assembling and accessorizing

The pi-top arrives in a box barely larger than the laptop itself, with the parts stacked in foam packing. It is accompanied by a manual that consists mostly of illustrations, all of them barely larger than the credit-card-sized circuit boards themselves. For this reason, the manual is best viewed online, where a few zooms can make the details clearer.

Anyone with hardware experience can probably assemble a pi-top in a hour or two. By contrast, the chances for rookie mistakes are numerous. Those can include everything from mistaking the screws in the kit for the spacers that position the circuit boards on the track to not realizing that what appears to be a full-sized SD card is actually a painted sheath from which to pull the actual micro SD card. Consequently, beginners might spend as long as four or five hours—longer if they drop the 2.5mm screws or nuts and have to scour local computer shops for replacements.

When assembly is complete, users need to insert the power cord so that the battery can charge. If the power button fails to light up within a couple of minutes, the most likely problem is a loose cable. Another possibility is that the Hub is not properly connected to the battery, and needs to be unscrewed and repositioned.

The pi-top runs Raspbian, the Debian-based distribution that has become a Raspberry Pi standard. Logging in offers a choice of a registered login, which is needed for upgrades from the company, or a guest account, which can only run applications. Given that the system is configured for sudo, and that both registered and guest accounts share the same customization, the reason for this login distinction seems obscure. Possibly the fact that a registered account is supposed to be able to login to any pi-top is intended as an aid in teaching labs, although as a feature it raises self-evident security issues. However, re-configuring sudo or replacing it with a regular root account should prevent such intrusions.

Either choice of account brings up pi-topOS, which is a shell for Raspbian with icons that link to four applications plus one for the LXDE desktop, and a configuration menu in the lower right corner. Pi-topOS seems targeted at younger users. The default links can bring up the command-line interpreters for Python and the Pi-targeted language Wolfram, as well as CEED Universe [YouTube], which is a gamefied approach to learning basic programming concepts, all of which reflect the original assumption that the pi-top would be used for do-it yourself projects and education. Only a link to a version of Minecraft suggests other uses.

From pi-topOS, users click into LXDE, where the default software is also sparse. However, those who wish to run the pi-top as a regular laptop can find most of the same applications in the Raspbian repositories that are available in any other Linux distribution, and can run apt-get to provide themselves with web and email applications, as well as LibreOffice 3.5—a somewhat antiquated but adequate version. Another option is to install from the limited selection at the Pi Store.

Running the pi-top as a standard laptop also requires additional hardware. The Raspberry Pi itself has only 8 gigabytes of storage, so anyone who wants a fully outfitted laptop may want to buy a micro SD card with more memory and install Raspbian and pi-topOS on it. Users may also want to add a flash drive or portable hard drive for additional storage.

A bigger problem, though, is that the keyboard connector and wireless dongle occupy two of the four USB ports on the Pi, so a USB hub is a sensible investment. Since the touchpad is to the right of the keyboard, left-handed people might want to add a mouse.

In addition, the pi-top has neither audio nor speakers. Audio could be added by attaching a speaker to the Raspberry Pi's built-in audio output, but sound has always been mediocre on the Pi—and besides, the pi-top is laid out so that extra cables are difficult to position. For this reason, an external sound card and speaker is probably the best choice for both convenience and sound quality.

The need for these extras seems another indicator that the educational goals were partly put aside due to the difficulties during production. Or perhaps another perspective is that, in a do-it-yourself laptop, users are expected to be resourceful when they customize. However, pi-top is aware of the need for sound, and announced a $15 speaker add-on as a stretch goal for the forthcoming pi-topCEED, which should work on the pi-top as well.

Educational goals

Like Raspberry Pi itself, pi-top has always intended its products to be educational. However, its first product shows only the first signs of this goal in a few default applications that are mostly undocumented.

Admittedly, assembling a pi-top is an education itself. Yet pi-tops are more simply laid out than a standard laptop or workstation, and the instructions fail to include very much that explains what most of the interior cables actually do. Instead, users are left blindly following instructions with minimal sense of what they are doing. Labeled cables and a handful of well-placed sentences in the instruction manual would go a long way to improve the educational experience of assembly.

To be fair, in the second crowdfunding campaign, education seems much more of a focus than in the first. Not only do donors have the option of donating a pi-topCEED for education, but the intent seems to be to provide more applications and tutorials for every level from basic electronics to programming languages; the first week's donors will be receiving a free breadboard and basic tools. If this material is delivered, then reviews of the pi-topCEED may be at least as much about the software as the hardware. But, for now, education is more of a promise than a realized goal.

Where the value lies

At $299, the pi-top cannot match the prices of Chromebooks and entry level laptops that are in the same device class. The pi-top suffers, too, from the need to add hardware, and the mostly undelivered educational tools. The impression is that the manufacturing process distracted pi-top, and that the company intends to regain its focus with its second product.

Rather, the appeal of the pi-top is in its ethos. Not only is it as free-licensed as any commercial computer is likely to be, but the idea of a laptop as customizable as a workstation scarcely exists outside of a few poorly-distributed kits.

In the do-it-yourself subculture around the Raspberry Pi, the pi-top offers a more portable controller than anything that currently exists. However, the greatest appeal of the pi-top lies its demystification of hardware and its encouragement of the view that a computer's interior is accessible and adjustable, rather than a black box. Although the assembly can be frustrating to those without hardware experience, the sense of accomplishment the first time the pi-top boots successfully is all the greater for the delays involved.

My own satisfaction in the process can best be summarized by this: although I was as clueless during assembly as anyone could be, the day after I got my pi-top up and running, I pre-ordered a pi-topCEED without any hesitation so I could learn more about computers at the circuit-board level. Frankly, I'm hooked—and judging from the fact that the pi-topCEED campaign reached its goal in two days, so, I suspect, are hundreds of others.

Comments (5 posted)