@timothyja: The FreeBSD Foundation might be worth a look - it was they who hired Konstantin Belousov to bring Kernel Mode Setting for Intel GPUs to FreeBSD, so they would probably be able to help with most issues that may arise.
SlaveBSD has its own proprietary masters! Definately NOT worth a THOUGHT.
Yep, the another option is that the big projects (Kernel, X, Mesa, KDE, GNOME, etc.) will collect funds and suggestions from users.
I can suggest to X, Mesa and Kernel teams that further development of opensource radeon drivers and I'd like to spend my money to that development. For example, there are 50 users that want further development of radeon drivers. We send our money to X, Mesa and Kernel teams and so creating the fund for radeon drivers. In turn X, Mesa and Kernel teams organizes a special contest like GSoC and XorgEVoC with special emphasis on radeon drivers.
Such a strategy is more flexible, I think. First of all, mentors from the big projects can formulate an appropriate task for each developer. Also they can control the development process and the quality of code. Also they can evalute the suggested project.
Yes, I know that donations are possible for (almost) every big project or even to special organizations (like SPI, http://www.spi-inc.org/), but I think that the existing donation system should be improved. Yes, I like to donate my money for X, but not for everything in X, I like my money to be spent for radeon drivers. And so on...
This "strategy" will fail, because you have X, Y, W, Z independent projects with untrackable people behind.
The correct strategy is single entity with clear goal that focuses solely on commercial open-source solution.
1) negotiates the GOAL with project X, Y, W, Z
2) with people VOTING on variants later on
3) publish required MONEY.
4) wait for coding offers.
5) check best variant (together with affected entity X, Y, W, Z) and give money to programmer.
6) submit the result upstream to (X, Y, W, Z)
If it becomes milestone-based then count me out. I wouldn't want my employer to pay me only when software was released or bugs fixed, so I don't think it's fair to ask somebody else to be paid like that. Expenses are regular (rent, food, electricity etc), so I think the developer's remuneration should be too. Unless the idea is that this is for people who already have jobs but want some extra cash.
Aside from that, wouldn't a milestone based system discourage somebody from refactoring code as necessary, incentivising them to rush something out that meets the milestone but isn't of such high quality? "Sorry it's ugly guys, I wanted to pay my rent".