alex chiang: web 6.0

I’ve recently discovered the Linux hater’s blog and think it’s brilliant. It’s like Fake Steve Jobs, but I actually care this time.

In the most recent installment, the Linux hater makes fun of the Foxconn conspiracy that wasn’t. I read the original thread and thought to myself, “what a wanker”. Look, I even said so on internal irc:

#hplu.25.log:25-07-2008 13:32:08 > achiang: what a wanker

The Linux hater says:

You may think that it’s totally lame that HP has to work around a Vista bug like this, but at least they can. They can count on “Windows 2006 SP1″ always meaning the same thing, and so they can fix problems for their users. The Linux kernel community has failed to provide stable targets for the rest of the industry to design around. They’ve made it extremely backwards and difficult for hardware vendors to fix things for Linux specifically, even if they wanted to.

As someone with a passing knowledge of ACPI, I see this simply as a pragmatic approach to the realization that Linux isn’t the market leader, and rather than fighting an impossible fight of begging resource-constrained hardware vendors to test with Linux, it’s better to simply assume the Windows code path is tested, and draft behind all that validation effort.

This approach is clever in at least one sense — Microsoft has to spend resources on their WHQL infrastructure, hardware vendors do QA against it, and then Linux gets to benefit. Of course, it’s a sub-optimal approach because Linux has to waste time figuring out how to be bug-for-bug compatible, but the important thing to keep in mind is that the world isn’t static. As lenb says:

Linux will continue to claim OSI compatibility with Windows until the day when the majority of Linux systems have passed a Linux compatibility test rather than a Windows compatibility test.

My prediction is that this future isn’t a complete pipe-dream, as hardware vendors attempt to squeeze more and more cost out of the system, the cost of WHQL certification increases as a percentage of the R&D. If the Linux market becomes big enough, it’s conceivable that a vendor decides to validate against a standard Linux test suite.

And while it’s madness to think of all the kernel/userspace/distro combinations out there, I note that the actual ACPICA in Linux doesn’t change very rapidly, so the vendor could test against a given ACPICA release, rather than any given Linus or distro kernel.