[Emaderton3]

I should add that we who use QNX in cars are mostly using it because we’ve invested so much in it already. It’s the incumbent choice. It’s what we know and it’s what works.

But again, new car companies like Tesla have no reason to use QNX at all.

So, just add BB10 to it then. Can't be that hard. ~

[Chuck Finley69]

So, just add BB10 to it then. Can't be that hard. ~

Plus there’s millions of BB10 fans that would pay for it.... /S!!!!!!!!

[Emaderton3]

Plus there’s millions of BB10 fans that would pay for it.... /S!!!!!!!!

Exactly. I can't imagine my car serving off the road as I am swiping the display off to the right of my steering wheel.

[Dunt Dunt Dunt]

Plus there’s millions of BB10 fans that would pay for it.... /S!!!!!!!!

Once it's 75% off the retail price, or they can get a hold of a used copy.... But if someone could hack it so we could install it for free, that would be preferred.

[Emaderton3]

Once it's 75% off the retail price, or they can get a hold of a used copy.... But if someone could hack it so we could install it for free, that would be preferred.

But of course it cannot be hacked because QNX is magic!

[co4nd]

That is due to the very time constraints you mentioned and it being EoL'd. Why would Google invest in building a microkernel if it offers no advantage in mobile? What is its expected return on investment that BlackBerry can not realize? How incompatible are drivers for previous generation SoCes? For example, if the code that identifies the GPU is replaced would the driver work? If so, how compromised would the performance be? On the desktop the GPU vendor wouldn't want a (previous generation) driver to compromise the customers experience and the associated bad PR, but here we are talking about good enough and being particular as to where to invest development dollars.

Google funds alot of projects that don't necessarily have any potential market advantage. Its founders are uber nerds who have the resources to dio anything they want. I don't think they're focused solely on mobile either. They are a company that will always be looking for the next big thing in information technology. mobile just happens to be it right now.

[DonHB]

But why did both companies choose to build microkernel based OSes? Linux is available and they chose the cost of starting from scratch. China has access to Linux like everyone else. BlackBerry has such an OS. Wouldn't it be cheaper for Google to buy QNX? Wouldn't BlackBerry sell it if the price was right? There must be an advantage to microkernels you aren't acknowledging.

[DonHB]

Linux can be configured in so many different ways. Even the scheduler is pluggable.

How hard is it to replace the memory management and the file system? How coupled is GNU to Linux?

Curious what tools you use for building on QNX?

[Chuck Finley69]

But why did both companies choose to build microkernel based OSes? Linux is available and they chose the cost of starting from scratch. China has access to Linux like everyone else. BlackBerry has such an OS. Wouldn't it be cheaper for Google to buy QNX? Wouldn't BlackBerry sell it if the price was right? There must be an advantage to microkernels you aren't acknowledging.

What does Google accomplish purchasing BlackBerry to get QNX and why, in regards to it’s current strategy?

[app_Developer]

What does Google accomplish purchasing BlackBerry to get QNX and why, in regards to it’s current strategy?

If Google did that, they would have a large part of the automotive market for core OS and core devtools for infotainment and instrument clusters.

What would they do with that? As a customer of QNX, if BB were acquired, I would still want our ironclad controls on the data from our cars. Because I want that, and I'm probably not alone, I think Google would not be interested in that business.

From a technical perspective, they would get absolutely nothing of interest. As I've said a few dozen times now, QNX was cool 25 years ago. There is nothing special about it in 2020, other than the fact that it is popular in cars (because car companies chose it a long time ago and it does the job just fine).

There was also nothing particularly special about QNX in 2010. Dodge managed to convince another Canadian company that it was something to build a phone on. He did that, sold the company, and then went to work for Apple. Good for him.

[DonHB]

So, technically, what is special about HarmonyOS or Fuchsia?

[conite]

So, technically, what is special about HarmonyOS or Fuchsia?

The former is for IoT devices. The latter is vaporware - but it is also designed to be scaleable down to embedded systems.

[app_Developer]

So, technically, what is special about HarmonyOS or Fuchsia?

Fuchsia is a project. If interesting things come out of that project, it will be used. I haven't seen anything earth-shattering yet. Dart seems the most interesting part.

Harmony has a massive political advantage: It's completely Chinese. As an emerging superpower, China wants more tech that they own and control. That's natural. It's what new superpowers always do in all of history. That's why they are investing countless billions into all sorts of tech. The work they are doing in ML/AI is staggering in scale.

[Chuck Finley69]

So, technically, what is special about HarmonyOS or Fuchsia?

You mean besides either project having access to practically unlimited economic resources as needed?!?

[DonHB]

Yes, what technical advantages have these resources delivered?

[app_Developer]

Yes, what technical advantages have these resources delivered?

Asked and answered above. Nothing yet.

[DonHB]

The former is for IoT devices. The latter is vaporware - but it is also designed to be scaleable down to embedded systems.

And why would Google not want to use this in mobile?

[app_Developer]

And why would Google not want to use this in mobile?

They will, if and only if, it gives them some advantage (business or technical).

If it gives them no advantage in phones, they won't use it. They tend to make pretty rational choices over there.

[DonHB]

Asked and answered above. Nothing yet.

That's what I wrote way back. That BlackBerry EoL'd the product before advantages could be realized. Any idea why Google chose to waste money on a microkernel architecture? They can afford risk, but even so, they must see potential there.

[app_Developer]

That's what I wrote way back. That BlackBerry EoL'd the product before advantages could be realized. Any idea why Google chose to waste money on a microkernel architecture?

As I said earlier, Google may realize some advantage on the development and testing side, at Google scale. Google Scale >>>> BB Scale.

Let's go back and summarize a bit:

If you are making a brand new kernel as a project, because you're a huge company that routinely spends billions on research projects, then you might want to start with a microkernel and then figure out how to make it execute as fast as Linux does. A lot of research kernels in the past 30-40 years have started as microkernels, including 2 that I know of at Microsoft, and the one that is used today at Apple.

OTOH, if you are a small phone company, and you throw away all of the free support and free drivers and all of the SoC support in Linux, just because you want a microkernel, then I'm going to call that stupid, because you didn't really get anything useful other than a lot of costs that you couldn't afford.

Incidentally, Microsoft dropped both of their microkernels because in practice it gave them no advantage. NeXT/Apple made theirs less micro-ey because they didn't see any advantage either (compared to the costs)

[DonHB]

I think NeXT/Apple could not get the performance it wanted. That is why part of the GPU driver in Windows is in the Kernel too.

[app_Developer]

I think NeXT/Apple could get the performance it wanted. That is why part of the GPU driver in Windows is in the Kernel too.

I assume you meant "could not get". That's right. Both kernels that I worked on in college and grad school were microkernels. Everybody starts with that design because on a whiteboard or a chalkboard it looks right, and it's really easy to get started on as a team, especially if the hardware you're working on is also co-emergent. Then practical realities set in and you have to adjust usually.

It's like Dr. Porsche when he started in cars. He made an electric motor that was in the hub. That is a nice simple design. But then he experienced the reality of the limitations of that and moved on to the more complicated, but much more realistic, flat-4 (and then f-6) that made him famous.

[DonHB]

But, in time, the performance issues were resolved.

[app_Developer]

But, in time, the performance issues were resolved.

In the case of cars, sort of. More than 100 years later. But there are still real limitations there. Cost is one. Also try to get one around any decent size race track. Or see the range on the autobahn. Or imagine the infrastructure problems if everyone in the world could suddenly afford one. Our newest all-electrics now are more performant, and we are making progress on multiple issues, but this is 100 years later.

And it's because we have a serious problem with petrol engines. Otherwise we wouldn't bother with any of it.

[DonHB]

I wonder if the need to be *nix like constrained what could be done with microkernels. So much hasn't been done due to entrenched tech (incl. knowledge) which makes the cost of replacement prohibitive.. IoT could be an excuse to rethink things more fundamental than the transition from the combustion engine to electric -- more like steam engine to electric engine.

We really are entrenched.