Note I described MNT reform as -mostly- open and the Precursor as -maximally- open (as in the the maximum extent currently possible to mass produce).

If Bunnis ASIC efforts succeed, then we have auditable reasonably fast chips in the next few years and a truly 100% open device. Tropic Square is another to keep an eye on.

Fully aware of everything in your descriptions here, but you always repeat this stuff as though I am not. Probably useful for others though.

Where we always seem to disagree is you usually try to dismiss mostly open solutions as no better than mostly closed as though the effort to pursue transparency is pointless. I feel every single component with open firmware and open hardware is a huge win, making accountability and community improvement possible. Likewise every blob is an eyesore that should be reverse engineered and replaced... or switch to more transparent alternatives when they exist.

Sure, auditing never catches all bugs, but it catches a -lot- of them. There are many severe security flaws I would never have had a chance in hell of having the time to find in closed binaries, let alone fixing them.

Sure underhanded C and all sorts of sneaky bugs can exist, but an open C solution could be replaced with an open Rust solution structured for east auditing or another language that makes it harder to do many common types of sneaky in.

If a vendor will not let me look at their code, I am extra suspicious of glaring backdoors or bugdoors until proven otherwise given countless examples in the wild.

I have always agreed open source alone does not mean code can be trusted. Most open source code is shit and should -not- be trusted (I review it for a living) but I am absolutely certain open source is an prerequisite to a community maintainable trustworthy solution existing where we get both freedom and security.

> Note I described MNT reform as -mostly- open

It's near completely closed source hardware. The SoC providing nearly the whole core system is fully closed source. An open source boot chain after the closed source early boot doesn't change this. Other components are closed source too. It's closed source with open source bits in between. Compared to the complexity of the SoC, radios, etc. the open source parts are insignificant. Open source between closed source components with most of the complexity it not mostly open source. It's simply not true.

> and the Precursor as -maximally- open

It's possible to use an open source RISC-V SoC instead of programming a CPU with a closed source FPGA. They don't use a closed source FPGA to be maximally open but rather to be closer to being able to inspect it.

> Fully aware of everything in your descriptions here, but you always repeat this stuff as though I am not. Probably useful for others though.

I don't think you're unaware of it. You must be aware the MNT Reform has a fully closed source ARM SoC with most of the core system's complexity but you still call it mostly open source.

> Where we always seem to disagree is you usually try to dismiss mostly open solutions as no better than mostly closed as though the effort to pursue transparency is pointless. I feel every single component with open firmware and open hardware is a huge win, making accountability and community improvement possible. Likewise every blob is an eyesore that should be reverse engineered and replaced... or switch to more transparent alternatives when they exist.

They are not mostly open solutions. It's false marketing. Open source does not have the properties you claim it does of heavily avoiding trust in the developers or providing much better security.

> Sure underhanded C and all sorts of sneaky bugs can exist, but an open C solution could be replaced with an open Rust solution structured for east auditing or another language that makes it harder to do many common types of sneaky in.

Memory corruption isn't required for serious subtle vulnerabilities and even safe Rust has plenty or room for memory corruption. Rust does not making auditing easy. It makes it easier than C, which is a low bar. Auditing C for deliberate vulnerabilities can easily be harder than auditing assembly code without anything that looks obfuscated.

> If a vendor will not let me look at their code, I am extra suspicious of glaring backdoors or bugdoors until proven otherwise given countless examples in the wild. > > I have always agreed open source alone does not mean code can be trusted. Most open source code is shit and should -not- be trusted (I review it for a living) but I am absolutely certain open source is an prerequisite to a community maintainable trustworthy solution existing where we get both freedom and security.

There are many glaring vulnerabilities in the most widely inspected open source projects including the Linux kernel. Many have persisted for not only years but decades. Open source does not inherently result in all these vulnerabilities being found and fixed, whether they were intentional or not. Open source can help with it but it provides no guarantee of better security.

Linux kernel is a typical collaborative open source project where performance, scalability and features trample over security. It being such an expansive and collaborative project means there's massive attack surface for intentional vulnerabilities and it doesn't have serious protections against it. Lack of prioritizing correctness and security for nearly all of it is pretty much equivalent to intentional vulnerabilities. Deciding not to deploy very useful features for finding / fixing vulnerabilities due to minor work it creates is typical, such as not marking intended overflows to have automatic overflow checks as an option. There's massive pushback against very basic things. The effort to introduce Rust for drivers has gone horribly despite lots of resources and it's face far greater resistance in the core kernel. Meanwhile, iOS has a kernel increasingly focused on security where they overhaul the whole thing for it. This is an example where one company controlling a project without collaborative is a massive win for security. There are projects like SQLite which don't take on the collaborative and open development aspects of open source. AOSP is similar to an extent, but heavily uses collaborative open source projects like Linux as core parts of it which largely don't have the same significant focus on security it grew over time. AOSP is about as security focused as iOS itself, but open source projects they use including Linux certainly aren't.