Left2MyOwnDevices

I would argue there is a substantial difference between the robustness of an internet server and the robustness of a kernel. But beyond that, the leap to safety critical systems bothered me.

I'd toss up a colorful and illuminating analogy, but better to just point out that the interfaces to safety critical systems are carefully analysed for vulnerabilities, and personally I'd have to be offered a pretty compelling reason to link 'outside' on an interface like TCP/IP.

It is a lot easier to build an interface using a deterministic command set with defined functions and be done with it.

This seems rather silly- no operating system is fundamentally secure, unless it is completely disconnected from the Internet. It makes no difference whether it is open or closed, proprietary or free. Linux APPEARS more secure than other systems for a couple of reasons:

1- Limited market share. Return on investment for malware developers is quite limited.

2- There are way too many variations to the basic Linux to make it practical to develop a "one size fits all" attack vector.

3- As a general rule (not an absolute, by any means), Linux users are usually more knowledgeable, and better prepared to detect an attack, and more likely to have appropriate protection in place.

Great points. Your assertion no OS is fundamentally secure is certainly correct. The three points you cite for Linux seem to be aimed at Windows. The trend I'm pointing out is there's more Linux being deployed as embedded, and that's putting some pressure on the RTOS guys to re-find their niche.

The point I'm questioning a bit: Linux more likely to have appropriate protection, in an embedded app? Enterprise, disk-based, IT people with access, no question. Patching the kernel regularly can be a problem in a flash-based system, or in systems not set up for OTA, or even in systems set up for OTA as our one Tweeter indicates - patches may be available but tough to apply. If the problem isn't in the kernel - say, it's in a contributed driver - who fixes it, in a timely manner?

That's what I was driving at as a problem with the "free" model. It recently took the community 2 weeks to get a simple fix to K-9 Mail, a popular Android app, when a version change stopped reading image types correctly. Two weeks. That could mean a lot of lost revenue if it stopped a product shipment cycle.

Not disagreeing, trying to get the other point of view represented fully - thanks again for your comment.