I think WebKit is the way to go just because the toolkit was already abstracted - the only issue is that it abuses GCC4 and C++ which makes it a lot more fun to debug.
As for the PROM, I don't know how much the source would help anyway except in the very specific case of the O2 where we have a pin and bus compatible CPU module with different initialization parameters. It wouldn't help much with porting third-party OSes since it's just bringup and runs before the OS anyway - plus, for most systems (Indy, O2, Octane with VPro, and increasingly Fuel and Origin3xxx) the reverse-engineering heavy work was done long ago, and what remains is just a lack of implementation. For example, BSDs have real accelerated CRIME support but Linux doesn't, and VPro is understood and we can render triangles on it but nobody wants to write an X.Org driver.
An open-source IRIX would be great if thousands of programmers with endless amounts of free time suddenly emerged from the woodwork to maintain it and add the features people want (USB, video drivers, and so on)... but for some reason I don't see that happening even if the source were available right now.
josehill wrote:You need a strong leader to avoid featuritis in open source software. Never mind that -- you need one in any software development project that involves more than one person.
I can't second this enough. Open-source projects with fairly strong leaders (Linux, at least until recently when things have been getting iffier) or corporate management (notice how WebCore/WebKit took off after Apple grabbed KHTML) tend to do really well. Open-source projects with only spirit and no strong leadership tend to either fall into disrepair and abandonment (too many to count) or a complete and utter mess (Firefox) depending on how popular they are.