Making schedulers runtime pluggable would make it really easy to get other people hacking on the Linux scheduler though.

For example, you could lash up a reusable test harness to allow scheduler testing under well defined and repeatable scenarios. This may then allow more direct comparison between schedulers hopefully leading to a best of breed race. Making it runtime un/loadable would also speed up the development cycle for the scheduler much in the same way that loadable modules can often be more rapidly debugged and fixed by not needing a reboot for each change.

You could even go crazy and make a scheduler plug in 'shim' just for the purpose of profiling different implementations under real workloads.

The only thing I would say is that the whole scheduler API should be made in such a way that the scheduler is undeniably covered by the GPL. Binary blob schedulers would be the worst possible outcome and would go against the thought of trying to open up the scheduler as a means of furthering development and healthy competition.

Lots and lots.

I find Java a lot of fun.

They probably worked out the costs if they sold >1M. They sold 100k, so never reached those economies of scale.

That's a shame, but at least they were thinking big. If they started out planning to sell 100k, they wouldn't have bothered.

I wish them luck in what they do next. Pleo is still unique.

> and of course, having a filesystem and a special MTD driver for
> *every single SSD drive manufactured* when they change flash
> chips or tweak the controller, could get unwieldy.

Large numbers of flash chips can be supported by the MTD CFI drivers:

http://en.wikipedia.org/wiki/Common_Flash_Memory_Interface

Something similar could be done for SSDs too, except they've chosen HDD standards as they are a better fit.

Mike

Yeah, well Via's driver support for Linux is horrible, while the Intel Atom stuff just works perfectly with Ubuntu out of the box. That may have something to do with why Intel is preferred.

Shame the TCP/IP stack isn't on the microcontroller. Putting uIP on there, or grabbing bits from my stack (http://www.mcternan.co.uk/MAD/) would be awesome.