follow me on Twitter

    Blog has moved to, see November, 2007 on new site

    Blog entries for November, 2007

    This post moved to

    For years now I've complained about the term "desktop" - an evil word that blinds us to more interesting opportunities to use free software in consumer-facing products.

    Look at all the recent examples.

    Built on GNOME technology, we've had Maemo from Nokia and OpenMoko for a while, and One Laptop Per Child.

    More recently, we have the Kindle and Android.

    Then no less than three consumer PC-like devices, running an Internet-centric flavor of Linux; Zonbu, Everex gPC, and Asus Eee PC.

    Zonbu emphasizes "Hassle-free. Guaranteed." because it backs up all your data to the Internet for $12.95 per month. The OS is completely stateless, i.e. doesn't keep anything locally that isn't backed up.

    gPC asks you to "Imagine an OS with easy access to the best Web 2.0 can offer" and shamelessly encourages you to think the "g" means Google as best they can without getting sued... "We recommend Google for just about everything... Gmail, Gtalk, Calendar, Maps, Docs and Spreadsheets, and more. We'd like to welcome you to the idea that Google already is your 'operating system.'"

    According to its press release, Eee PC "is focused on providing users with the most comprehensive Internet applications based on three Es: Easy to learn, work play; Excellent Internet experience and Excellent on-the-Go."

    Compare to the concept for Online Desktop proposed at GUADEC (incidentally before any of these three had come out, or at least before I knew about them):

    The perfect window to the Internet: integrated with all your favorite online apps, secure and virus-free, simple to set up and zero­ maintenance thereafter.

    There's a real problem with Zonbu, gPC, and Eee PC: they are all running one-off, hacked-up software that's specific to the hardware. This can't last. If this type of thing catches on, eventually there's significant consumer benefit if the software is "hardware independent" and there's a relatively stable platform used by as many people as possible. Consumers benefit from using a platform lots of other people are using. It's also helpful, of course, if there's a thriving, upstream open source community behind the platform.

    The opportunity for a project like GNOME is to ignore proprietary, legacy desktop operating systems and focus on huge, unique advantages:

    • Stateless / zero-maintenance
    • Works with the Internet - all the apps people are interested in - not locked-in to Windows Live
    • Runs well on small, low-power hardware that is green and portable
    • Common platform and data, spanning mobile and home devices

    It's pure idiocy to chip away at matching Windows or Mac feature-for-feature, hoping to get from 90% of the feature matrix to 95%, wishfully thinking that will matter. "Linux will be ready for the desktop when these 5 pet peeves are fixed" magazine articles drive me nuts, because they assume such a terrible slug-out-the-feature-matrix strategy. Letting the incumbent define the playing field is suicide.

    The right approach for the free software community is to offer a different product, and what's changed in the last few months is that no less than three companies have shipped first rough cuts at what I'd consider the open source appliance of the future. While at the same time, One Laptop Per Child shipped an even more innovative PC to kids around the world, Nokia continued to ship the Maemo mobile environment, and Google piled on with its own open source mobile platform. There's a lot happening.

    If I had magic fiat power over GNOME or Fedora or Ubuntu or whatever, these are the opportunities I'd be interested in.

    There's nothing really new here; after all, the general idea is not too far from what I assume Eazel had in mind. Before that, there was the i-Opener (IMO a good product, but Netpliance messed up the business aspect by selling at a loss without locking people in to a contract).

    The idea is old but today's world is new: increasing use of broadband, decreasing hardware costs, more capable web-based apps, better mobile devices, and open source software improvements have made open source client appliances more timely and realistic today than they were a few years ago.

    It will be interesting to see which of these efforts succeed, and which companies and projects end up at the center of gravity. I feel very confident in predicting, however, that if a free software OS gets in front of a substantial number of consumers, it will be in the form of these new and different products, not in the form of a strictly traditional desktop operating system.

    This post moved to

    I wrote a short Red Hat Magazine article, complete with screenshots, touring online desktop in Fedora 8.

    This post moved to

    Apparently someone has decided to be pedantic and set my font DPI to 50 by default. Without changing some number of other things, this is a regression, and broken. To avoid the many problems this creates the decision was made long ago to simply make the DPI always 96. This is also what Windows does, I believe.

    If people want to be pedantic, they need to solve the problem that on a perfectly normal laptop, when I log in by default, the desktop looks terrible until I go into the Advanced tab of the Font dialog and fix the DPI to be 96 again.

    This is not something to be documented with some FAQ on the wiki about going into the Advanced tab. It's something that needs to be fixed by default.

    I have no idea of the history, but I bet some theory about "scalable user interface" was involved here. Maybe that idea makes sense someday, but for the screen sizes most people have today, the desktop looks bad unless it's drawn mostly pixel-by-pixel. It's cool to play with scalable UIs, but it needs to be done while continuing to use the pixel-by-pixel code for traditional screen resolutions.

    (I'm assuming Fedora 8 inherited this; though hopefully we patched a fix into the spec file to override busted upstream decisions, or will soon in an update if it doesn't get fixed upstream. I'm testing with a jhbuild from GNOME svn right now.)

    (And no, I don't care about any counterarguments unless they involve a near-term patch that makes the desktop look good by default again.)