Can Linux Dominate Smartphone OS?

Jeryl Kesh writes "Does Motorola's roaring success with its Linux-based 'Ming' phones in China indicate that the open-source platform is now a serious contender against Symbian and Windows Mobile in the handheld device software platform arena? The world No. 2 mobile phone maker, which debuted the Ming smartphone in March this year in China, shipped more than one million Linux-based units in China alone last quarter, according to research firm Canalys. However with Nokia refusing to adopt Linux, Symbian remains by far the top mobile device OS, according to Canalys, with a 67 percent share, well ahead of second-place Windows Mobile, with 15 percent of the market. Eirik Chambe-Eng, the co-founder of one of the most popular mobile Linux platforms, Norway-based TrollTech, has also reportedly predicted a 'revolution' in the use of open-source software on phones and handheld devices. His contention was that Linux gives handset manufacturers and OEMs (original equipment manufacturers) 'complete control,' and in turn keeps Microsoft and Symbian at bay."

Comments

Linux demand is growing

[gillbates]

My company markets reference designs for handheld devices (typically cell phones, media players, etc..) to OEMs. Our customer demand for Linux has increased dramatically in the past year. We've doubled the size of our group, and we still can't hire engineers with Linux experience fast enough.

Yes, we do WinCE development too. But, the WinCE group not only has at least twice as many engineers, they are also behind the Linux group in terms of features. When it comes to rapid development, there's simply nothing better than Linux, because most of the work has already been done. This allows us to concentrate on adding features that differentiate us from the competition, rather than on merely getting something working.

But WinCE also places substantial roadblocks to rapid development. A routine build of WinCE takes 20 minutes; a clean build takes more than an hour. By comparison, our average Linux build time is about 30 seconds, with a clean build taking about 15 or 20 minutes. But it gets worse for WinCE:

1. The FAT32 filesystem is a major liability for embedded devices. Because of the fact that the disk head must seek back and forth from the filesystem table to the actual data, the effective data bitrate decreases with time. This means that WinCE has a maximum practical encoding time of about 1 hour; after that, the filesystem driver just can't keep up. We don't have this problem when using ext3 under Linux.
2. WinCE doesn't have a native terminal; you have to recompile and reload the whole OS and application image in order to test a change of even a single line of code. Worse, you can't interactively debug the board because you have no way to send something to standard input.
3. The WinCE API is relatively new compared with that of Linux/UNIX. Our customers do not need to buy expensive documentation packages from Microsoft in order to work with our embedded Linux solutions; their engineers already know the Linux API, and can begin work immediately. When one considers the fact that the average consumer electronics device has a saleable lifetime of 3 to 6 months, development lag time becomes a critical factor.

Quite frankly, I'm glad to see the demand for Linux growing. However, I'm also putting in quite a bit of overtime because of it, so it is sort of a mixed blessing.