Hacker News Reader

Comment by shakna | Hacker News Reader

> Our goal is to ensure that we get all aspects of our products mainlined so that custom kernels will not be required.

I love these kinds of projects, but after being burned by the PocketCHIP, who also said this but never made it there, this is isn't somewhere I want to go again.

View on HN

The big problem with the C.H.I.P. is that it used raw NAND flash as the main storage (and MLC NAND on the original boards at that), which just wasn't something that could be supported upstream. I think more or less everything else required for upstream support was there a long time ago. This should hopefully avoid that problem; it's using EMMC which doesn't have that problem and it looks like it's based on the Allwinner A64 which already has fairly good mainline Linux support.

Yeah, ditto. Why can't mainline Linux support come first? Then they could sell a working product.

Crazy, I know. Investing in difficult R&D problems to produce a quality product is not a good business model. It's so much easier to churn out a working devkit and ask for pull requests.

Sorry for the snark and pessimism, but that's what happens when you keep getting burned. This statement does not inspire confidence:

>Finally, to ensure that Pocket P.C. has great software support out of the box, we have requested embedded development help from the community.

Why is software always an afterthought with these projects? Especially a project like this, which has clearly seen an enormous amount of effort put into the hardware design.

> Why is software always an afterthought with these projects? Especially a project like this, which has clearly seen an enormous amount of effort put into the hardware design.

This can go either way, in my experience. I’ve worked with software developers who do a terrible job with hardware, and hardware developers who do a terrible job with software. Best I can tell, it’s a problem that’s existed since we started making computers.

In some ways, I think it’s arrogance. The rockstar software developer believes that software is infinitely more complex than the hardware it runs on. The ninja EE believes that software is trivial. And they both assume that they can effectively do each others’ jobs.

The only example I have is the LibreComputer S905X based board - it took years of work with a small team to get the kernel changes upstreamed. Hell, look at the Raspberry Pi - I think mainline support is pretty recent.

This project shows a lot of promise.

> TRYING TO BUILD THE ULTIMATE RASPBERRY PI COMPUTER (ZERO TERMINAL V3)

https://n-o-d-e.net/zeroterminal3.html

https://www.youtube.com/watch?v=wiJqUWfR90I

I suspect if someone were ready to pounce on the Oxide source code when it comes out, that one might make a go of a properly open portable handheld device by adding some drivers and snipping a few bits off. Radios might be the hardest things to sort out, since people seem to like to keep those as proprietary as hell, and WiFi is merely okay now.

Sounds like starlink antennae won’t be small enough for a long time, so that leaves who besides Qualcomm, Broadcom, and Apple?

Mediatek make modems

Oxide?

Brian Cantrell is off trying to build server hardware based on Open firmware. Which likely means they’ll have to write a lot of it themselves.

Reply on news.ycombinator.com