FAA Grants RSC Status to Linux-Friendly RTOS

BoulderDad writes "LinuxDevices.com is reporting that a proprietary RTOS capable of running Linux binaries has been certified by the FAA as a re-usable software component (RSC). LynuxWorks says LynxOS-178's RSC acceptance will enable greater software reuse among integrators and developers of safety-critical aerospace and defense components."

Comments

OMFG!

[Anonymous Coward]

That's a lot of acronyms!

NGTH

[Kawahee]

The article says it allows for better integration into mission critical applications. However, I don't see this happening.

Realistically, mission-critical developers aren't going to trust code written by the public, certified or not. There's no responsibility to the developers if something goes wrong with that code.

Re:NGTH

[Chirs]

Of course you don't trust it. That's why you review the code and make sure it looks okay.

Even that can be a whole lot cheaper than writing it yourself.

Re:NGTH

[Valar]

You know, there is an actual vendor selling this, right? As in, there is a company that sells it, that you could go to if something goes wrong.

Re:NGTH

[JesseMcDonald]

I hope that you weren't implying that LynxOS-178 was "written by the public" -- the summary and the article were both indicated that LynxOS is a proprietary RTOS capable of running binaries compiled for Linux. Despite the name (LynuxWorks), the system is not derived from Linux in any way.

Re:NGTH

[Fly]

I think he must be referring to the applications, not the OS itsef. LynxOS is not Linux. It's proprietary real-time OS that can run Linux applications. The LynxOS itself is backed by the vendor, and it's pretty good from what I hear. However, the applications built on it depend on the skill of the application developers, not the OS vendor.

Re:NGTH

[WindBourne]

Actaully, certification is all you really have. To obtain OS-178B is very difficult.

Microsoft was approached by my company to get OS-178B. Once they looked at what it would entail, they called back a week later and told us that they had a good laugh. In their own words, not even Vista will come close. And XP was not even a consideration.

Be sure to read the article. This is LynxOS with Linux API on top. That is much easier to do.

But if you check google, you will find that there are several other companies with OS-178B version of Linux. They are a pain to work with as they are nothing but a stripped down redhat with a few re-written parts. Do you think that before I write code for any of these, that I am going to check over all the code? Not one line. I trust that the FAA and the company that sell these did that already. Why do I do that? Because, I do not have the time to do that and write my code.

That is why we use certificated OSs in critical areas of the cockpit.

Re:NGTH

[WindBourne]

Derived? No. Use the same code that Linux uses? Yes. The core kernel is mostly their work. But the API and all the supporting tools are mostly from GNU/Linux.

Re:NGTH

[Alioth]

Navigational systems are fairly critical things (especially when IFR). I was slightly shocked to see that a friend's Apollo panel mount GPS ran on Windows NT 4.0!

Re:NGTH

[PPGMD]

Huh? Ghee thats why we have GPS approaches?

Just about anything permanently mounted to an aircraft requires FAA approval, most early GPSs were not IFR approved, but now almost all panel mount GPSs have certification for enroute navigation, and many have approval for approach use (on GPS approaches).

I know this for a fact because I had a field inspector yelling at me about a camera mount until I showed him that it was removable, and not a hazard to flight.

The FAA
We're not happy, until you're not happy.

Re:NGTH

[WindBourne]

FAA Aproval is nothing. There are various classes and certifications of instrumentations. I am a developer, so I do not really get into all that stuff, but here is the general breakdown;

• Class A; a laptop that you carry with aviation equipment or a GPS.
• Class B; an instrument that is IN the dashboard. But all it gets is POWER. It is not allowed to interact with anything else.
• Class C; In the dash and ability to read the data from the aircraft bus; that is it can display the status of the aircraft.
• Class D/E; in the dash, and not only reads, but writes data on the bus; that is it can be used for control.

If anybody else has the real scoop, go for broke on it.

Oh, great.

[Mr. Roadkill]

/dev/altimiter not found
GE-xxxx: scsi2: AEN: WARNING: SMART threshold exceeded: Engine #3
Kernel panic: defect on /dev/wing/left - printer on fire?

Acronym overload

[Life700MB]

* FAA.
* RTOS


--
Superb hosting 20GB Storage, 1_TB_ bandwidth, ssh, $7.95

Re:Darn Acronyms

[misleb]

If you think IT uses a lot of acronyms, aviation is 10x worse.

Want a weather report?

KPWK 202253Z 04015KT 10SM SKC 01/M06 A3018

-matthew

Re:Darn Acronyms

[tlhIngan]

It's not just acronyms. It's mixed units. A METAR in North America (Canada, at least) will get you the temperature in degress Celsius, windspeeds in knots, visibility in statute miles, and cloud bases in feet. (We'll leave the altimeter setting as mmHg as a side issue.)

Of course, TAFs are worse. And lets not forget the shorthand for weather conditions (rain/showers/etc) comes from French.

Acronyms

[Digitus1337]

I for one welcome our new acronym... OL's.

Re:Actually. . .

[Sqwubbsy]

Actually, anagrams are when you create a word or words out of another word or words' letters.

And pricing...

[sadr]

And a little research turns up per-developer pricing, although not the per-unit run-time license cost. That's not actually unreasonable, given the cost of DO-178B Level A documentation, but still. Ouch.

Price and Availability
In addition to the LynxOS-178 kernel, the offering also includes a complete artifacts package for the kernel and user library, DO-178B required documentation, code coverage test suites and analysis for 100% modified condition/decision coverage of the kernel and libraries, a full suite of standards-based development tools, and support. The company will also soon release the industry's first commercial-off-the-shelf certifiable TCP/IP stack. Development seats, including the LynxOS-178 kernel and one year of priority support, start at $18,000.

Re:Actually. . .

[AusIV]

No, acronyms are abbreviations that form pronounceable words. Examples are LASER and RADAR.

Initialisms use the first letters of words. Examples include WTF, OMG, and the things in the article.

Anagrams are words that are made by rearranging the letters of another word: Clint Eastwood -> Old West Action, Mother in-law -> Woman Hitler.

There were no acronyms or (intentional) anagrams in the article, just a bunch of initialisms.

But cash strapped developers ...

[Ungrounded Lightning]

And a little research turns up per-developer pricing, although not the per-unit run-time license cost. That's not actually unreasonable, given the cost of DO-178B Level A documentation, but still. Ouch.

Note that, because it's a Linux API, the bulk of the development can be done on Linux platforms WITHOUT per-developer licenses.

You'd need occasional testing against the real OS by someone "sitting in a licensed seat" - to check the behavior under the real OS's scheduling regime and detect reliance on missing or divergent features. And of course you'd have to hammer on it ifn licensed seats (and real or excelently hardware modeled aircraft devices) for final test. But if the licenses are sufficiently dear you concevably might end up ahead. (You wouldn't need per-seat licenses for initial prototyping work, either.)

(The "reliability tested in later" nature of such an effort wouldn't be an extra burden if machines connected to prototype hardware or timing-accurate models of them also aren't available at all seats all the time.)

A lot of software might not need close modeling thoughout development to get right.

Less worried about /dev/altimeter

[jd]

It's when the plane becomes overcrowded and the OOM Killer starts deleting passengers that take too much room that you might have to be concerned. Or selinux is enabled and the pilot doesn't have the right security label for the brakes...

Back to the LynxOS stuff, though. If LynxOS can run Linux binaries, then people can develop on Linux and run under LynxOS. (Duh!) As the hardware for development is orders of magnitude more expensive than the development tools, I'm not sure it'll have much short-term impact in that direction. HOWEVER, it may result in top-of-the-line developers for aviation software migrating to Linux for basic development, which may pull some more of the commercial sector in that direction, as those developers HAVE to have money to burn. It may also result in bug reports from a new set of power-users, as the additional stresses reveal problems that more conventional usage isn't exposing. That may lead to improvements in Linux that wouldn't otherwise occur.

It would be nice if LynxOS could do the same thing SGI and IBM did eight to ten years ago, now, which is to release kernel code fragments that people could experiment with and adapt into Linux or one of the BSDs. (Yes, they both did filesystems too, but I was thinking more of SGI's OB1 code release - an open-source set of Orange Book B1 security modules. I don't believe anyone ever used the code, which I think was stupid, but I feel confident that enough people learned from it that the security enhancements in Linux and the BSDs today are further along than they would have been.)

It would also be nice if the few aviation electronics companies that produced Linux drivers either updated them (Linux 2.2 is old and wasn't the most stable series anyway) or they should Open Source them. If nobody can use the drivers as they are, it's pointless to have them on the website. If the drivers are free downloads anyway, it's impossible for the company to make a loss if someone were to produce a driver that worked better. And if someone DID produce a driver that worked better, the company might sell more hardware (either with a big stack of indemnities, or a higher pricetag to cover the re-certification).