I'm fairly certain that's not accurate. Most in-dash units are based on QNX and use their own dev platform.
Thank you! Ford just kicked Microsoft out after 2013, and they were really already out, just left their "last bastion." QNX is not open source, although some various kits are for it, and open source still has to be certified.
I.e., open source does
not automatically mean "better." I'm the first one to admit this, and can cite countless examples.
SIDE STORY: Just read up on the private lawsuits against Toyota, where several lawyers (usually with a BSEE in engineering and lots of embedded experience), have reproduced and proven their problems in court (including how breaking becomes impossible due to a combination of factors). Most of them have revolved around how Toyota's management decided to use a non-certified open source stack and devkit, as well as source boards without watchdog timers and other things, different than design by the engineers (who used certified boards and certified open source tools and platforms). It explains why and how Toyota engineers actually did have certified platform, including good code (even control code verified by NASA), but the hardware changes and use of non-certified tools and platforms as a cost-saving measure in production, is where their problems occurred.
Whether I rebuilt RHEL SRPMS at or using TimeSys/MontaVista builds, or compacted a RHEL run-time for various customers (before and while with Red Hat), it's still not a full, certified release ... and much less if I use any old build. Embedded is a mess all on its own, but as Michael Tiemann ran into when running Cygnus (the first, profitable open source company), especially after Red Hat acquired them (Cygnus was bigger at the time, and actually making money, but Red Hat went IPO), there's just not a lot of margin in the space. So everything comes back to Time'n Materials, making the continual investment, getting certified solutions for specific industries, etc... Hence where I came in on-loan to Global Engineering Services (GES) regularly, even working with them before I was hired direct mid last decade.
Microsoft doesn't like to stick around in this space for the same reason, although they sure do spend a lot of money trying to keep it going in select industries, in the hope they don't lose them entirely. Because that affects more than just the embedded, but the backend -- where the real money is at -- more than the front-end. So that's why they do it.
A lot of the new stuff is HTML5/JavaScript.
Exactly! I had this argument in 2002, and had to revisit it in not only 2009, but again in 2014. I had a customer who kept writing in VBA, then IE6 and now .NET in the last 5 years they cannot even port to the latest version, and have to scratch. At some point, just sticking with open standards (putting open source aside) gets a heck of a lot easier, and saves a lot of money ... and time!
Nothing is flawless, but anyone who keeps regurgitating, "Oh, Microsoft really means it this time," is usually just listening to marketing, while ignoring all they've conceded on RT, ARM and Windows 10 being the last version. As I said, they are making great moves on the Server with Nano for Windows Server 10. But they still keep playing games on the client/library front, which really limits their opportunities. Their run-time is a joke, and it's going to get people killed in various uses, if they push it there yet again (like they did years ago with 911 systems).
I think even Microsoft/Ford Sync (which was based on an outdated version of Windows CE) used native code for everything and not .NET Compact.
Yep. .NET continues to be limited to non-real-time, non-mission-critical, non-safety, and only when Windows interfaces are used, and legacy can be avoided. But a lot of their developers write in Visual Studio, and focus only on the desktop, which no "target" module/plugin can completely absolve (despite some 3rd party claims to the contrary with their products).
I've been in those rooms where I had to tell them I cannot port something because they used all sorts of libraries that just don't exist in Mono, let alone we'll have to run the WINE run-time just to emulate various things. I've also been in the rooms where they've been fed the marketing, and we go through the platform and library-level details. As always, I say the same thing ... "If you're Windows, just run x86 and be done with it. Don't try to get off of it ... at least not today" (and likely not until after the non-Windows successor to Windows 10, given the recent announcements).
Microsoft basically abandoned their connected car platform. It wasn't worth porting from WinCE to NT. This is the real reason why Ford ditched it in favor of QNX.
Microsoft tries to "prop up" a "last bastion" in various industries. But at some point, they just give up because of the burn rate of subsidy, and -- more importantly -- they are just hurting their own image.
In retail, it was Best Buy in the early '00s, although .NET is regaining some mindshare, I'll fully admit, as you can get low-power x86 now. But retail has limited requirements for safety and real-time, and finances are reconciled daily. But early on, they were basically funding all of Best Buy's retail. I know, I had to deal with this first-hand in the late '00s, as
THE guy in Red Hat doing embedded retail appliances. Customers are always about how much they can get Microsoft to concede, and pay at a loss, and when they don't, they went elsewhere -- usually with 100s of developers kicking and screaming "job security."
In trading, it was the London Stock Exchange (LSE). Even when it was tanking, the same day Fannie and Freddie went under and they were down virtually all of it, Microsoft blamed the integrator, Accenture. Accenture was being entirely funded by Microsoft, and the LSE was heavily subsidized without or with very limited cost to the LSE. After losing that, they re-focused back, entirely on the desktop, mainly non-real-time dashboards and status. I mean, in 2001, NT and Linux 2.4 were almost on-par in response times, especially with FSMLabs licensing their patent to Microsoft and implementing a solution similar to RT/Linux for NT. But then you had TimeSys (who I also worked for awhile) make real-time almost commodity in 2.4, and so forth. By 2.6, it was over.
Bloomberg and Reuters are the top 2 in the securities software industry, and do use .NET, because they are focused on desktop info solutions. #3 in the industry overall is (or was) IPC Systems, who develops most of the real-time stuff interface stuff (yes, for users, not just backends), first VxWorks (IQ/MX), and then Linux (IQ/Max). I worked on the IQ/Max backpack in Fairfield, CT back in '06-'07, when IPC was trying to make themselves sexy for a buy-out (which happened). The last company I ever worked for that still did even their PCB fabrication -- not just design, but even fabrication -- in the US.
Auto ... well, already covered, Ford was the last bastion. Scion, who has completely different electronics from Toyota, has been getting "beat up" with their return rates because they are using a lot of fly-by-night open source platforms and devkits from what I've seen. I'm also the
first person to
openly complain about Android's sustainment, and plenty of my colleagues who are "Google apologists" have also heard from my other colleagues that they understand where I'm coming from. But Android itself is psuedo open source, not a GNU platform, and not exactly Java one either (again, the new Microsoft -- I actually agreed with
some of Oracle's arguments on that front).
I have just dealt with a lot of .NET code over the years, including recently, written entirely from developers who cannot think outside of the default Visual Studio targets, and it's a joke. Microsoft is trying to address it with the stacks, but when you get into it, it just sucks so bad, because they cannot get their own tool developers inside the company to write good tools for anything but desktop x86. That's always been the problem, hence the strategy, which is really a crappy one. They got into Mono too late, and so much is just x86-only libraries, which is why Windows has to come to an end to solve it (otherwise, they never will, and they've conceded this).
It's always just being pushed back, because it's their cash cow they cannot lose.
He had some sorta absurd security add-on that was blocking the JavaScript or something.
Yep. Just habit after getting a client-side compromise years ago (my NIDS caught it, otherwise I would have never known). I fully admitted it was something I was overlooking, and I asked people to post a screenshot so I knew what to look for (tags are difficult to find when you don't). Once I saw it on-screen, I was able to find another tag near it, and that was a quick fix.
But I know, I'll never hear the end of it.
