Crave

Google's Android software gives Sun Microsystems' Java technology a starring role--but not the version of Java the rest of the mobile phone industry has been developing since the 1990s.

Instead, Google struck off on its own in an attempt to improve performance and openness for the software used in the Open Handset Alliance phones. That means programmers will have a new variety of Java to reckon with--offset somewhat by Google's $10 million code contest to draw developers in.

One difference is Google's development of its own core Java virtual machine (JVM) technology called Dalvik, the software that actually executes Java programs on an Android phone, which Google says means Java programs run fast even on the constrained hardware of mobile phones. But a more significant departure than just using an in-house JVM is the fact that Android isn't part of the Java Community Process that Sun established in 1999 to oversee the development of new Java features.

The JCP governs Java by codifying new features as application programming interfaces (APIs), so programmers can have a standard way of calling upon new technology such as Bluetooth support or 3D graphics. But that existing Java realm wouldn't accommodate the developer freedoms Google thought were important in Android.

"We wanted the platform to be open in a lot of different ways," said Mike Cleron, a Google senior staff engineer working on Android. "The idea is that anybody can come along and replace the pieces of the Android experience on a very fine-grained level. The existing APIs didn't really allow the level of openness we were hoping to achieve in Android."

It should be noted that Google isn't working in a Java vacuum. For example, one of the OHA partners, Motorola, has helped lead development of Java for mobile devices, and Google wants to keep the Java programming experience familiar to developers. And Google is an executive committee member of the JCP, though only for the Standard and Enterprise editions that run on PCs and servers, not the mobile edition for phones and other devices.

"We have people on the team who are active in the Java community. They've been helpful in informing us and guiding us, making sure what we were doing is familiar to folks in the Java community," said Steve Horowitz, Android's engineering director.

Further fragmented?
But the bigger issue is whether Google's effort will worsen the already fractured world of Java. Not all phones support all the same Java standards, so programmers can't be sure that their software will run on a multiplicity of devices, as the "write once, run anywhere" Java tagline promises.

"They are using Java, but they aren't implementing any well-known Java framework, and really that just creates another standard to support. The risk they take here is that they might fragment the market further," Benoit Schillings, Trolltech chief technology officer, told my comrade Maggie Reardon. Trolltech, which sells tools and components for programmers whose software runs either on PCs or on mobile phones.

Mauro Lollo, CEO of mobile phone video-streaming company Movidity, saw Google's work similarly. "In essence, they've created another standard. Standards are great, but the challenge is that there are so many of them," he said.

Google also faces a common risk of open-source software, that the openness will mean programmers can "fork" projects in different, incompatible directions. (Indeed, this was one of the earlier reasons Sun resisted its eventual decision to make Java open-source software.) "In the end, you could have 20 different versions of the Android technology that are incompatible, because anyone can take the license, modify it, and create another variation," Schillings said.

For its part, Sun supports Java and open-source software on mobile devices, but expessed some caution about joining Google's alliance. "We were interested in being part of the Google ecosystem, but we were interested in getting more clarity on what this program entails," said Rich Green, executive vice president of Sun's software effort.

Asked if there's any possibility of unifying the Android work with the Java Community Process, Horowitz said, "It's an open alliance. We can welcome anybody who wants to join."

Android uptake
Techno-politics aside, Google clearly has grand aspirations for Android. And it wants outsiders to be part of the development.

In stark contrast to Apple, which plans to release a software developer for its iPhone in February, half a year after the product began shipping, Google is releasing its SDK about a year before any Android phones ship.

"We're making it available pretty early--early enough that we can get feedback at a point where we can still impact the direction of the software," Horowitz said. "People tend not to ship SDKs until the products are done. In this case we thought the platform was such an important part that we wanted to get that out early."

Of course, there's another advantage to releasing an SDK early: the open-source community can help build interesting applications that give Android phones more than just the basic set of programs.

So far, so good, said Horowitz, pointing to "unprecedented" interest in Android compared to other projects hosted at Google's open-source projects site, code.google.com. "It is above and beyond anything Google has seen to date," Horowitz said.

A diagram of the inner workings of Google's Android software for mobile phones.

(Credit: Google)

Among details in the SDK:

• It makes mention of support for GSM mobile phone networks, the leading technology for mobile phone networks, but is silent on support for the top rival, Qualcomm's CDMA. That will come, though, Horowitz said, pointing to CDMA allies such as Qualcomm that are members of OHA. "It's clearly something on the roadmap, but we're not talking about specific support for it at this time," he said.

• OHA supports touch-screen technology, but Horowitz declined to comment on support for multitouch, a notable iPhone ability that opens up user-interface possibilities, beyond saying multitouch support isn't in the first version of the Android SDK.

• Google will release a new version of the Android SDK once feedback from programmers starts coming in. "We're committed to a regular release cycle," Horowitz said.

• Software should run quickly on mid-range phone hardware such as those with a 200MHz ARM 9 processor. "One of the key goals of the project was to ensure we can run on a broad range of phones that don't require a high-end processor at all," Horowitz said. "When we bring it to higher-performance devices, it's just going to scream."

• The SDK so far permits development only of software that runs on the Java foundation, not natively on the hardware itself. "We are aware of the interest in native application development, but we having nothing to comment on right now," Horowitz said. But performance shouldn't be an issue: "Our system is designed to take full advantage of native code in performance-critical areas and expose this functionality through our framework APIs."

Marguerite Reardon and Dawn Kawamoto contributed to this report.

Correction: An earlier version of this blog misstated Google's connection to the JCP. Google is a member of the Java Community Process, though not for the Java Mobile Edition version to which the Android software is most closely related.

Forgive me if I appear a little skeptical here about Google's Open Handset Alliance. By my count, it's the fifth consortium so far to attempt to craft something useful for mobile phones out of Linux and open-source software.

OHA has by far the highest profile, it's got the most persuasive list of members, and its timing is the best. What's not yet clear is whether the "Android" work of Google and its allies will unify or further fragment work in the area.

Rallying programmers behind a unified effort could help determine whether this effort will accomplish more than the Linux Phone Standard (Lips) Forum, the Open Source Developer Labs' Mobile Linux Initiative, the Consumer Electronics Linux Forum (CELF), and most recently, the LiMo Foundation begun in 2006. Related efforts one step removed include Intel's Moblin and, Nokia's Maemo, and any number of other open-source projects.

Just as with PCs, somebody has to write a "stack" of software spanning from basic operating system functions all the way through communication utilities, user interfaces and Web browsers. Unlike PCs so far, though, the mobile phone market has suffered from a profusion of incompatible software foundations, despite some efforts to use Linux and Java to bring some common ground.

... Read more

SAN FRANCISCO--One area where Sun Microsystems' Java caught on was in mobile phones, but a leader of the project is working to eventually replace the mobile-specific version of the software.

Sun Vice President James Gosling speaks in May at the JavaOne conference in San Francisco.

(Credit: James Martin/CNET News.com)

Java Standard Edition (SE), geared for desktop computers, will gradually supplant Java Micro Edition (ME) as technology improvements let more computing power be packed into smaller devices, said James Gosling, the Sun vice president often called the father of Java.

"We're trying to converge everything to the Java SE specification. Cell phones and TV set-top boxes are growing up," Gosling said at a Java media event here Wednesday. "That convergence is going to take years."

The prime example of the trend is Sun's own JavaFX Mobile, software Sun got through its SavaJe acquisition and which the company hopes mobile phone makers will embrace. JavaFX Mobile includes almost all of Java SE, though it's missing a few pieces such as CORBA (brace yourself: Common Object Request Broker Architecture) for getting software to work with other programs across a network.

Sun's Java expectation dovetails with recent trends, most notably Apple's iPhone, which architecturally is much more an Apple computer writ small than a mobile phone writ large. In particular, Apple uses a version of its regular Safari Web browser so users will have as much of the desktop Internet experience as possible.

At the same time, Intel is working to bring x86 processors that run PCs into mobile gadgets. It's in cohoots with open-source efforts including Ubuntu Mobile and Mobile Firefox .

The move to Java SE won't happen overnight. Rich Green, Sun's executive vide president of software, said he expects smart phones using various pared-down versions of Java to stay in the market for at least a decade.

But the shift already was under way. "All the work in Java ME had been pushing it closer and closer to Java SE," Gosling said.

Defragmenting mobile Java
Moving to Java SE could help fix one nagging problem with Java ME: fragmentation.

Java ME is a collection of abilities--basic ones and higher-level options layered on top--each defined by a detailed description called a Java specification request. For Java ME, there are a large number of these JSRs for various features. That posed a challenge to Java's original tagline, "write once, run anywhere."

The tagline came about because a program written in Java could in principle run on any computer that had a Java virtual machine. The JVM is a software foundation that lets a generic Java program run on a particular computer. But with the multiplicity of Java ME extensions, there was often little guarantee that a program written for one mobile phone would work on another.

Java SE has a much richer basic set of abilities, so using it instead of Java ME could at least in principle restore some of Java's promise of software portability.

JavaFX mobile is one component of a multipronged effort called JavaFX that Sun announced in May at its JavaOne conference.

"JavaFX is probably the largest and most complex software engineering effort Sun has ever done," Gosling said. Here's a quick tour of the JavaFX components:

Tour de Java FX jargon
Unless you're a serious Java nerd, and maybe even if you are, Sun's latest nomenclature is a crazy hodge-podge of terms. Java SE--OK, that's been around for nearly a decade, we can handle it. Though there was some numbering madness a few years ago, Sun seems to have settled on the current version being Java SE 6. But let's work outward from there.

First comes Java 6 Update N, formerly called the Consumer Java Runtime Environment (JRE). This is an attempt to make Java SE easier on the average computer user, chiefly through improvements to the plug-in that Web browsers use to deal with Web pages using Java.

Among the Update N features: It preloads Java when the computer boots to avoid the excruciating delay when you encounter a Java Web page. It installs faster by loading only a bare-minimum kernel--typically less than 4MB--that gets things started and then updates itself with the full 12MB Java software collection. It takes advantage of Windows' Direct3D graphics abilities. And it includes a more graphically modern user interface that gives a unified look across multiple operating system.

Update N should go into beta testing in December and be available a few months later, said Chet Haase, Sun's Java SE client architect.

Atop Update N comes JavaFX Script. This is a new scripting language geared specifically for fancy user interface actions such as transparency and other effects that are difficult with the prevailing Web browser scripting language, JavaScript (which contrary to what its name may imply isn't based on Java). JavaFX Script is geared toward use more by design types than engineers, Gosling said.

Of course, you can't have a script without something to understand it. Thus there's JavaFX compiler to translate people's code into instructions the computer can execute.

Last is the aforementioned Java FX Mobile. This software is in part a reaction to gripes by Java ME developers who wanted a more unified foundation, Gosling said. Another difference compared to Java ME is that Sun will deliver it as a prewritten binary program; Java ME typically comes as source code that programmers must compile into something useful.

Potshots at the competition
Gosling and Java have been at the vanguard of an idea that in a way is just coming back into vogue: rich Internet applications, which is software that runs in a Web browser but comes with a lot more pizzazz and capability than bland Web pages.

Java caught on as a way to run server software and to run games on mobile phones, but one original promise of Java was turning a Web browser into a foundation for sophisticated software. (If you're having flashbacks to Netscape taking on Microsoft Windows and the resulting federal antitrust case, just breathe deeply for a moment to settle down.)

But much of the rich Internet application action is happening with software such as Ajax, the Adobe Integrated Runtime (nee Apollo) and Microsoft's Silverlight and Google Gears.

Gosling thinks JavaFX has a chance, too, though, listing several advantages he believes it has: a richer user interface, faster performance, a robust and well accepted language and better abilities when a computer is disconnected from a network.

And security, he adds. Adobe's AIR is designed to let programs work like regular PC software, but Gosling thinks the approach unwise. "It's a petri dish for viruses. Security is really hard to implement well."

Another star is coming into alignment in the mobile Linux galaxy: Firefox.

Mozilla has set up a group to develop the Firefox Web browser for mobile devices, hiring new staff and elevating the priority of the work to the same level as desktop computers. Mike Schroepfer, Mozilla's vice president of engineering, announced the mobile Firefox move on his blog Tuesday evening.

"We are serious about bringing the Firefox experience and technology to mobile devices," he said. "Bringing Firefox add-ons, the Mozilla platform, open source, and a large and passionate community to the closed and fragmented mobile platform will do the world some serious good."

Schroepfer announced two new hires. One is Christian Sejersen, who recently led browsers at mobile browser developer Openwave, will be in charge of the mobile Firefox work and will set up a research-and-development center in Copenhagen, Denmark. In addition, Brad Lassey joined Mozilla from France Telecom's research-and-development group.

Mobile devices have become a "tier one platform set for Mozilla," he added. "This means we will make core platform decisions with mobile devices as first-class citizens."

Don't expect instant results, though. Mobile Firefox won't arrive until "later...certainly not before 2008." It will employ technology that will ship after Firefox 3, he added. That version isn't even in beta testing yet.

The work dovetails neatly with several other projects for open-source mobile devices. Canonical is working on a version of Ubuntu Linux for mobile devices, Intel trying to improve Linux for x86 chip-based mobile devices and includes Canonical as a partner in the effort, Nokia runs a project called Maemo for its Linux-based Internet tablets, and Google apparently has its own mobile Linux work under way.

While these efforts are marching in the same direction, if not necessarily in lockstep, it should be noted that mobile Linux efforts have been under way for years with little major success. One thing that's different this time is that Intel is working to bring the power of a fairly modern PC to small devices, potentially making software development easier.

Apple's iPhone has put an emphasis on full-fledged browsers on mobile devices instead of the limited-function ones that so far prevail in the market. "The user demand for a full browsing experience on mobile devices is clear. If you weren't sure about this before, you should be after the launch of the iPhone," Schroepfer said.

Apple's products use its own browser, Safari, but another option is Opera, which has had a long-standing mobile browsing effort, with products including Opera Mini and Opera Mobile.

But full browsers present hardware challenges for tiny devices. "Getting a no-compromise Web experience on devices requires significant memory (at least 64MB) as well as significant CPU horsepower," Schroepfer said.

The mobile Firefox project will replace another Mozilla effort, called Minimo, Schroepfer said. "While we don't currently plan to develop that project further, it has already provided us with valuable information about how Gecko (Firefox's page rendering engine) operates in mobile environments, has helped us reduce footprint, and has given us a platform for initial experimentation in user experience," Schroepfer said.