An interesting thread has been winding its way around the blogosphere lately, about something called the “Firefox operating system”. It got started when Todd Bishop offered some speculation about the mysterious venture that Firefox’s Blake Ross and Joe Hewitt are collaborating on: are they building a Firefox operating system? Dwight picked up the thread, but didn’t follow it. Ed Bott grabbed ahold long enough to show that the thread doesn’t lead anywhere.
Nobody says exactly what a Firefox operating system (let’s call it FOS) is supposed to be, and so I’ve been wondering about it. A Google search doesn’t turn up much, aside from the original speculation cited above, and others picking up the thread to weave its speculation into their blogs. The top hit belongs to a pseudonymous poster at spreadfirefox.com, who asks:
Why not make a Firefox operating system? Make it like MS Windows and Linux combined. Kinda like the ultimate OS. If that actually happens you can start selling the OS to companies like HP and Dell so they could attach the OS with their comps and make it successful.
(Some light editing by yours truly on the quote above.)
Clearly this poster is talking about a general-purpose operating system, and I presume that the others are too. What form would it take? I can think of three:
First, it could be an operating system written from the ground up. Ed Bott dismisses this idea, saying “It takes years to build an OS kernel from scratch.” Not only that, but it takes a lot of expertise in some specialized fields, and I’ve seen no indication that Ross and Hewitt are involving these kinds of people. But the best argument against writing a completely new operating system is that it just doesn’t make sense.
That leads me to my second theory: FOS could be an operating system built on top of an existing kernel and perhaps a few system libraries. For example, you could take the Linux kernel, the Free Software Foundation’s Glibc, and a handful of other libraries, and build just about anything you wanted to on top of their solid base. You could also build upon another kernel such as CapROS, which is an outgrowth of the Extremely Reliable Operating System, a novel concept in OS design. Or you could build upon OpenBSD and have the basis of an ultra-secure operating system.
Such an operating system needn’t look or work like Linux, even if built upon the Linux kernel. A kernel is just a very low-level set of services; the look and feel of an operating system depends more upon its high-level applications.
You would still have to provide a windowing system, a set of application support services, and maybe even a new filesystem, but by building on top of an existing kernel, you could eliminate a lot of the most complex and difficult work. If Ross and Hewitt are building a general-purpose OS, this approach would make more sense for them than writing from scratch.
My third theory is more mundane. Take a current Linux distribution and strip it of everything that makes it look or smells like Linux. Bundle it with a small, focused set of applications (Thunderbird, for example), and give them all a consistent, unified look and feel. Throw out 99% of the application-level cruft that has grown up around Linux, and reduce the operating system to little more than a support system for Firefox. This would go a long way towards eliminating some of my complaints about modern Linux distributions: that they have too much crammed into them, and that they’re trying too hard to be Windows.
Finally, I have one more theory that’s more radical. It’s surely not what most people mean when they say “Firefox operating system,” but it’s fun speculation nevertheless. Instead of building a general-purpose operating system, why not build an operating system specifically for Firefox? Here’s how you would do it…
Start with the Linux kernel. (I keep mentioning it because it’s free, open, reliable, and has loads of drivers in it.) Strip out anything not absolutely necessary to run Firefox; reduce the kernel to little more than basic process management plus some network and video drivers. Build a lightweight windowing system on top of this stripped down kernel. Next, write any application-level libraries that Firefox might need, or rewrite bits of Firefox to eliminate the need. Finally, glue it all together, and run Firefox directly inside this OS as a kernel task. (If a web browser, Tux, can run in the kernel, why not Firefox?)
If you’re thinking that all this Frankenstein-like merging is blurring – no, erasing the line between Firefox and the operating system, then you’re absolutely right. That’s the whole purpose: to merge Firefox and the operating system together so that they become one.
What would be the point of all this? Well, if Firefox is the operating system, then the computer boots directly into the browser, with just a short delay to initialize the network and video drivers. In other words, booting would be fast. A computer with FOS loaded onto it (a “Firebox”?) now becomes more like a TV or a radio: when you need it, you flip on the switch, and in a couple of seconds it’s ready to go. When you’re done with it, you flip it off.
With Firefox integrated into the OS, you can now run Firefox on much less machine that you would need for a general-purpose OS. You could probably get by very easily with a 1 GHz CPU and 128 megabytes of RAM (we’ll eliminate the Firefox “memory leak” somewhere along the way). By storing FOS in flash memory, and by modifying Firefox to store information such as bookmarks on the Internet, the need for disk drives is eliminated. OEMs could build special purpose systems with FOS preloaded and sell them dirt cheap. If a $100 laptop</a is possible, why not a $100 Firebox?
This kind of device isn’t a computer, not in the current sense of the word. It’s an information appliance for Internet use. If that term sounds familiar, perhaps it’s because I’ve used it before, in June, when I wrote about Bumptop. In that post, I mentioned user interface design guru Jef Raskin and his “humane interface.” Raskin designed an information appliance, the Canon Cat, that was made by Canon and sold for about $1500 apiece. The Firebox is just a modern renovation of the same idea.
Would a Firebox be able to compete against general-purpose computers in today’s marketplace? Maybe. Is this what Ross and Hewitt are building? Don’t be silly! It’s just a wild idea on a Saturday morning.