[immibis]

Any one of us here could learn the skills to design a smartphone. It won't necessarily be good, but I remember that years ago, someone made one with a touchscreen hat and GSM hat atop a Raspberry Pi, rubber-banded to a power bank. I'm sure any one of us HN users could do this. And it worked. Quality only goes up from there.

The problem is it won't run any apps, so you'll need to carry this open-source secure phone in addition to your normal phone.

[fsflover]

This is not as simple as you're saying. Making a new phone not relying on proprietary drivers tied to Android is impossible without a huge effort: https://news.ycombinator.com/item?id=21656355

[immibis]

Obviously you'd only choose hardware that works the way you want it to.

[fsflover]

Hardware relying on free drivers is almost non-existent in the mobile world. There is nothing to choose from, obviously.

[immibis]

Then aim for freely distributable drivers. You can share copies of Raspbian, so it seems possible.

[butvacuum]

You mean the Linux distro that exists because it needs to contain broadcom drivers/blobs/etc that are under NDA?

[fsflover]

Then your hardware will turn into e-waste as soon as the vendor decides so and stops updating the drivers.

[zdc1]

Or use everything via the web browser; but yes, I think apps are the main reason we can't just have a generic Linux phone OS on an open hardware platform

[bossyTeacher]

Apps make or break operating systems and app stores. Just ask Microsoft (Windows Phone) or Huawei (HarmonyOs). IIRC amazon was paying devs to publish to their app store or something like that.

Thankfully, some apps have both web and native mobile versions but for a modern digital life, the critical apps are sadly not on both versions.

[fsflover]

We have generic Linux phone OSes: Mobian, PureOS, postmarketOS and more. They can even work as daily drivers on some phones.

[amelius]

Isn't there an emulator that can run Android apps inside any Linux distro?

[immibis]

No. There are a few that claim to, but none of them are actually any good. Waydroid, for instance, requires that your kernel is compiled in basically "Android mode" (e.g. binder enabled).

[amelius]

How do the Android developer tools run Android apps on Linux then?

[immibis]

Inside a virtual machine which is easy to detect.

[yjftsjthsd-h]

> Waydroid, for instance, requires that your kernel is compiled in basically "Android mode" (e.g. binder enabled).

Waydroid needs you to have a single kernel module, which is in mainline Linux and just happens to be disabled in many desktop builds. That hardly makes it an "Android mode" kernel, and I certainly see no reason why it should make the system no good.

[immibis]

We can have it. It won't become as popular, but we can have it.

[mschuster91]

> Any one of us here could learn the skills to design a smartphone.

Unless you're Fabrice Bellard who literally created a 4G softmodem - no. It takes a whole lot of people (or, again, one genius Fabrice Bellard clone) to design a smartphone. You'll need AT THE VERY LEAST:

1) a SoC that has reasonably open device drivers and specifications - without that, all attempts are moot

2) a hardware engineer to deal with the PCB

3) a low-level system engineer to deal with the initial bringup (aka, porting u-boot and maintaining it)

4) an RF engineer to deal with the black magic that is designing ultra high performance PCBs that deal with the RF stuff (2G-5G phone networks, BT, WiFi, NFC, GPS) and high-frequency buses (storage, RAM, baseband, USB, PCIe, CSI/DSI)

5) a GPU driver engineer of the class of Alyssa Rosenzweig to get the GPU drivers to behave (she literally provided better-compliant drivers than Apple)

6) a battery engineer to ensure you don't end up with something like the ill-fated last Galaxy Note (that had to be fully recalled due to battery issues)

7) a ton of software engineers to get the basic things running that people expect from a smartphone (e.g. phone calls, 911, SMS, MMS, a browser and enough userland libraries so that third-party developers can begin to port games)

8) hosting engineers that deal with reliably delivering OS updates, application updates and A-GPS data

9) a skilled purchase and finance department to acquire all components as well as skilled QA people to make sure you don't get screwed in your supply chain by someone cutting corners or trying to engage in outright fraud

10) plastics and metal design engineers for the housing and other related engineering, and you'll probably also need engineers specializing in mass production and assembly as injection molding is a skillset on its own

11) engineers specializing in low power domains to get something that doesn't eat through the entire battery in a matter of hours

12) UX, UI designers to get something people can actually use (partially, that's also compliance stuff - think of accessibility laws)

13) testers to test your device against an insane load of other things - headsets, headphones, consumer and enterprise wifi, car head units, mice/keyboards, game controllers, USB hubs, monitors, projectors, adapters, dongles, IPv6 in its various abominations, phone network-side vendors, how devices behave in trains, cars, airplanes, cruise ships, in temperature and humidity extremes, under water, in back pockets (bending!), in dirt, dust, rain, being drenched in all kinds of beverages, muck, snow, fog, right next to extremely powerful broadcast radio transmitters, high magnetic/electric fields, teeth both human (toddlers) and animal (cats and dogs)...

14) logistics experts to deal with shipping, returns, refunds, recalls

15) customer support

16) psychoacoustics and acoustics engineers to make sure your device doesn't sound like shit (both what you hear, and that includes safeguarding the speakers from burning out, and what others hear from you, aka the beamforming stuff that the Asahi people reverse engineered)

17) video/colorspace engineers to make sure the whole darn thing isn't off color

18) camera/optics engineers, even if you acquire camera units these need to be integrated properly

19) lawyers and domain experts to deal with the compliance crap: RoHS, CE, FCC, India's regulatory authority, licensing, binary blobs, video codecs, audio codecs, carrier compliance testing, HDMI, HDCP, the RF compliance crap that's needed for US compliance [1], tariffs, sanctions laws... the list is endless

20) advertising (although admittedly, word-of-mouth could be sufficient), and PR in general (including websites, print media, AtL/BtL marketing)

21) deals with app developers, lest you end up like Windows Mobile

22) security testers/experts to make sure your devices don't get 0wned by cellebrite, mossad, nsa, cia, ...

23) human resources experts ("people engineers") to herd all the cats

24) packaging engineers to make sure the product arrives at the customer's hands both looking appealing and undamaged (tbh, that's at least four distinct skillsets as well)

You're looking at a minimum of 2-4 million $ for the engineers alone, another 4-5 million $ for the compliance crap, many millions for the app deals and way more in upfront cash for components and logistics chains.

That's why every attempt at a reasonably open source phone design has either failed or is many years behind the mass market. And the list of organisations attempting to do so include household names of the likes of Mozilla. And that is also why/how ODMs exist... they all have figured out some "minimum viable design" that gets tweaked a bit for the customer brand, and that's it. Everyone else went bust. Including, as mentioned, Microsoft. Including former powerhouses such as HTC. It's simply too complex to keep up.

On HN, we could probably drum together people of all these skillsets, no doubt (it took me half an hour to think of all these people and I'm pretty certain I've missed important aspects still!), and even ones with enough money to burn. But even then: the competition are the richest companies on the planet: Apple, Google, Samsung. Good luck...

(And yes: a minimum viable phone - probably a lot of people here including myself could whip that up using a COTS 5G modem, a Raspberry Pi and a power bank. But that's a MVP, not something you can sell to anyone less nerdy than Richard Stallman, and it's based off of the work of a lot of the people I just spent 58 minutes to think of and write down)

[1] https://github.com/lenovo/lenovo-wwan-unlock

[immibis]

or you could slap a GSM shield on a Raspberry Pi.

[mschuster91]

As I wrote: that's a MVP, not something you can sell to anyone less nerdy than Richard Stallman, and it's based off of the work of a lot of the people I just spent 58 minutes to think of and write down.

[immibis]

Then you can copy the relevant parts of the designs from both boards onto one smaller board.

[mschuster91]

Good luck getting that to perform. Or even boot. High-speed buses are the darkest of the dark arts to design, at the kind of frequencies particularly PCIe runs even the slightest trace length mismatch, impedance issue with the PCB itself, vias, or even external RF sources can, do and will mess with your sanity.

That's my entire point. It's not easy to design a complex thing such as a smartphone.