It's more complicated than that. If Google just produced a standard version of Android with a standard kernel version that all vendors would write for, refreshing and wiping phones would be as easy as Windows/Linux laptops. I wrote a post on this a while back:

http://penguindreams.org/blog/android-fragmentation/

Since then, I've become more and more frustrated with other ARM boards. Android fragmentation is directly related to ARM fragments and it's a huge mess. Things like device trees help, but ARM is missing the standardized architecture that Intel, Power, et. al. have.

> If Google just produced a standard version of Android with a standard kernel version that all vendors would write for

Then you wouldn't get all location data, call logs, and application usage data sent to China, you'd get them sent to the US, via Google, for better user experience and more targeted ads.

The result would be literally the same.

If you are in the US, at least your data isn't being sent to another country. Outside the US, I think your comment stands on its own.

Indeed I am outside the US, and I'm surprised the comment got that many downvotes.

EDIT: At the time, it was at -3

One approach would be to define an ABI-like interface above the kernel layer which all Android userlands would adhere to. It would specify the functions for libraries like libhardware and libstagefright, as well as the GLES or Vulkan extensions needed to be supported. Google would takeover software updates against this certified base, with the OEM providing the base system for their hardware.

are you sure the problem is arm? How can google use a standard kernel version? I mean it is quite standard and only changes between major releases doesn't it? The thing is that linux being monolithic requires all the drivers baked in, which is fine like it is fine the whole flashing ROMs stuff, or maybe it was a few years a go when phones where phones and not pocket pcs so they benefited from the whole packaging. Magenta by google seems to use userspace drivers, so if that is ever gonna get used in phones it might be a solution for a more standard kernel

Linux modules are very kernel specific. If you have modules compiled against the same kernel version, but with slightly differently flags, you won't be able to load it and will get a lot of invalid symbols.

There are thousands of kernel variants between every Linux distro. You always need to recompile drivers. nVidida and AMD/ATI get by this by providing binary blobs and a shim layer than can compile and link the two.

In the Windows/Mac world there is one version of the kernel officially maintained and distributed by their respective companies. They can keep the kernel consistent for a given release. That's what you have one set of drivers for Windows 7, one for 8 and one for 10, instead of a different set of drivers for each and every service pack.

Android could, at a minimum, standardize the kernel for every Android version. With that, binary drivers could work across devices.

Yes, the problem is ARM.

How do you want to boot your kernel? There is no BIOS or UEFI or other standardized firmware, that would initialize at least one CPU, RAM, storage and load your kernel from there. Every ARM implementation boots in different way.

After you boot up, how do you enumerate devices that your kernel can use? There is no PCIE or other plug and play bus, that can enumerate attached devices. Your OS has to know what is available. Poke a wrong port and your board freezes. This is the reason, why Device Tree is a thing.

Linux kernel does not require all drivers baked in - kernel modules existed for decades already. Monolithic does not mean that everything is one big blob. NT is also monolithic, for example.

I don't see what the monolithic kernel has to do with it. Kernel modules are dynamically have been loadable. Yes, they run in kernel mode, and yes their version tightly bound to the running kernel -- but how is that a problem for abstracting of varying hardware?

The strange thing is that I have less people around me with virus problems, stolen card problems, scam problems, etc than 10 years ago. Despite the fact that we deem those devices less secure.

Something is off.

I guess a lot of people have moved to platforms which are essentially virus-proof, like the iPad. I have several members of my family who used windows computers before, now they use iPads exclusively for all their work and pleasure. Previously any of them would open anything that they got in an email, and I guess they still do - it's just that with some devices, it's not an issue. A zipped up virus won't do any harm on an iPad, and probably most Android devices are secure from that too.

So while I think that windows devices are as easy to infiltrate as ever, a lot of people have simply stopped using them, and replaced them with - coincidentally, not intentionally - more secure platforms.

Pretty much this, security isn't improving, its just those infected zip & doc files usually aren't targeted at Android or iOS.

From trivial to apply app and OS updates to code signing and disabled sideloading, the mobile platforms (either one) make it really much harder to fall prey than on desktop OSes, by design. Malware authors have to hit a homerun on 0-days to even begin to get a foothold on the mobile OSes† these days.

† on properly updated devices such as the Apple ones as well as Nexus+Pixel. Probably this is where some form of vertical integration starts to really matter in practice.

Maybe criminals are more subtle now. For example most people who own devices that are part of the Mirai botnet are not aware of it.

I think you're right. Unless you're running crypto malware there's hardly any reason to let your victim know you've infected them. In fact it's best if they never notice anything is amiss.

It's a common pattern with a lot of "awareness" and media problems. There's a big problem with something, people start noticing it and "raising awareness" of it and publishing articles, books, etc on it. Now several things happen at the same time - people who are responsible for the systems that have the problem start addressing it and it gradually starts getting better in response to the publicity and complaints, but at the same time, the community of people finding and drawing attention to the class of problems gets bigger and develops more infrastructure and institutional culture. Groups get created and funded, people start full-time careers with those groups, regular magazines and websites get published and develop their own staffs, etc.

Usually, the underlying issue is complex enough that it takes years or decades to make the problem like 95% better. But the community around the issue has become entrenched, and won't go away just because the problem is basically solved. There's plenty of excuses at hand to justify the continued existence of the group - the problem is never 100% gone, so there's always something new to address, and they can always say that we have to be on guard for the problem coming back. But the effect is the same - the noise gets louder as the signal gets weaker.

If you're paying attention, you can see this pattern play out for all sorts of things.

In this case, we have both made the computers and browsers a lot more secure, and a lot of consumer-level computing activity has, for a bunch of reasons of which security is certainly one, moved to drastically more locked-down devices. Good luck building a botnet of iPads and Android phones. Many of the hacks that get the most publicity are super-complex one-offs that are in theory highly dangerous, but in practice are so valuable that they would only ever be used by state-level actors against handpicked national security targets, because you have to spend millions to find them, and they will become worthless about a week after the appropriate companies become aware of them.

Security on desktop OSes is still much worse, with even sandboxing being fairly rare.

Oh come on. The virus situation on Windows XP was appalling. The system was wide open. It's the primary reason I mainly switched to the Mac in 2006. Sane security was the one truly good thing that's come out of the post-XP era of MS operating systems.

Desktop OSes still run every app as the user, which gives them access to everything useful by default. Mobile OSes differ in this respect since they sandbox by default.