Linked by Eugenia Loli on Tue 26th Mar 2002 20:24 UTC, submitted by Jean-Baptiste Queru
Syllable, AtheOS Bill Hayden did the obvious: He forked AtheOS (which is technically similar to BeOS) and used its app_server and Interface Kit (without the use of X11) and rest of its kits on top of the 2.4.x Linux kernel. While the AtheOS kernel has some very nice features, by being modular, semi-microkernel, with good preemptive/multithreading support etc., it lacks a solid VM and swap support and of course, it lacks a good driver support, things that the Linux kernel provides. Bill Hayden accounced his fork on the AtheOS mailing list and made known that the "Atheos API has been merged with the BeOS API, there is PowerPC support, gcc 3.0.X compatiblity and OpenTracker/Deskbar as the desktop manager".
Permalink for comment
To read all comments associated with this story, please click here.
Re: Binary drivers
by Rayiner Hashem on Wed 27th Mar 2002 16:21 UTC

It's bad.
>>>>>>
Leave it to a simpleton to take all the subtlety out of a complex, two-sided arguement...

For all intents and purposes, virtually impossible.
>>>>>>>>
Keyword: Virtually. If a developer is going to do the customer enough of a disservice to provide only binary drivers, then he should have to work at it. The key question is: Is good support for binary only drivers worth the hassle to kernel developers? Since there's not many binary-only drivers out-there, and Linux has been pretty successful in getting companies to open their specs (Matrox, 3dfx, ATI) I'd argue that it isn't worth it to support binary-only drivers well.

First you say bad-bad, but instead of supporting your claim, you point a counter-example.
>>>>>>>>
No, I say its bad, but can be legitimate in certain cases, such as NVIDIA's. If you want examples of why its bad, I'll oblige you. You pay for hardware. You own it. Companies (through proprietory drivers) should not control how you use that hardware. Microsoft made me upgrade to Win98 just so I could use their stupid USB joystick. On Linux, people just backported the USB code to the older OS (kernel 2.2) so people could use USB devices without upgrading. The key idea is that people don't need to be slaves to the driver support of the company whose hardware they use. Second example. Drivers often have bugs that take a long time for companies to fix. It reassuring to be able to fix those bugs yourself instead of waiting for patches from the company. And yes, I have done this. The BeOS USB driver didn't work properly with my Precision Pro joystick. I went in and fixed it myself in a couple of minutes. It took more than a month for an update to be released that worked properly. Open standards are the way of the future. Just look at USB. It's very well documented, and an entire class of devices (HID, which includes joysticks, keyboard, etc) are based on an open spec. Without these specs, you'd only be able to use these devices on Windows.

My oh my, aren't you presumptuous!
>>>>>>>
Nope, just practical. If ATI and Matrox didn't lose by making their graphics card specs open, if Intel didn't lose by making UHCI and EHCI open, if the dozens of companies that provide open specs for their hardware (http://www.linux.org/hardware/components.html) don't lose anything by making these specs open, why would the developers of a MIDI device lose anything?

Thank you very much for providing me the arguments why the Linux kernel is NOT the way to go for a desktop OS!
>>>>>>
The Linux kernel is just a piece of technology. On a windows machine, you have no contact with the windows kernel. On a BeOS machine, you have no contact with the BeOS kernel. All the stuff that people don't like about Linux (GUI, configuration, complexity) is entirely in userspace. Using the Linux kernel doesn't do anything but base an OS on a proven, open piece of code.