I'd have gone down an optimization rabbit hole, while never finishing the original project.

Source: https://www.theregister.com/2023/08/29/freebsd_boots_in_25ms...

After asking the owner of said screen how he could stand that... "stand what?"

Yep, I guess most people are not that picky.

But more seriously... I don't think that Linux has ever been booted on a non-monolithic CPU (I wanted to say 'discrete cpu' first, but there's some PDP-11s with 4 chip CPUs)

Some points:

* there's a real lack of quality, up-to-date documentation. I would have thought that at least on Linux you'd find some documentation, but most of it seems to be "RTFS".

* BLE is in general very unfamiliar to most developers. There's no client and server, there's central and peripheral. GATT profiles are a mix between TCP connections and binary REST-ish interface.

* Encryption/authentication is possible, but depending on the manufacturer's API/quality of documentation it's not really apparent a. how to select a secure connection method b. how to even check if and which authentication/encryption was chosen

* Coming from the previous point, many BLE devices have the same generic GATT profiles, sometimes with the same sample data. This looks like a lot of BLE devices just copy&pasted sample code from the manufacturer and added the minimal changes "to make it work"

* It's probably really easy to do passive/active fingerprinting to find out the manufacturer and/or chip version used in a device. Default services, ordering of advertising options etc

* Many BLE devices are not conformant. Uninitialised name fields with garbage in them ("Device Name: WHOOP\020��=u5״\023n"), manufacturers using random identifiers that clearly don't belong to them

* when doing passive BLE sniffing: the biggest obstacle isn't getting data. It's how to filter it. One of the most useful filters of the nRF Connect app for android is to filter out all advertisement packages for apple and ms devices, to cut down the overwhelming amount of such devices

[1] https://fuglede.github.io/llama.ttf/

Coincidentally, if anyone knows how to figure out which Intel CPUs actually support 5-level paging / the CPUID flag known as la57, please tell me.

Nmap used to have a mode where you could portscan via a third-party FTP server.

And it just so happened that HP Laserjet printers had an FTP server active by default that allowed PORT to any IP.

But if you're totally comfortable with vim you'd better use ex, which basically is both an extension to ed, and the non-interactive part of vim.

Other fonts... I could see myself being distracted by changing fonts in a document, except maybe for comment blocks. But for those italic/thin seems to work well already.

Tried to find the tool... it's GNU enscript. Syntax highlighting for several languages, outputs to postscript.

Ouch.

I broke the space key on my dads computer while trying to get a new speed skating record.

TL;DR: Oklab is pretty simple, but is already pretty nice as a perceptually uniform color space. Darktable UCS takes Oklab and tries to reduce the residual error.

Feel free to correct me if I got anything wrong

As it stands now, the front page is pretty useless if you never heard of ghostty. It's a ghost in a terminal? And there's tty in the name.... so is it a program to show ghosts in your terminal?

I'd love to hear more about the low power design. You mentioned going for a MSP430 on here, and the high level architecture blogpost has some infos about the power domains too. But I'd love to see another blog post with some more nitty-gritty powersaving stuff.

I was also wondering... have you thought about adding a simple streaming compression algorithm to increase battery life? Less data to write would mean less time of VDD_B being active. Could maybe even be implemented on the FPGA. I was thinking of something like delta encoding, with two different code lengths for small and big brightness changes (per color channel.

E.g. first bit 0 -> 5 bits of +/-32 brightness delta following. First bit 1, 11 bits of brightness delta.

You'd loose some minimal information if brightness changes by more than 2048 between two pixels, but you could just "smear" that change over two consecutive values (e.g. [0, 0, 4096, 4096, 4096] would end up as [0, 0, 2048, 4096, 4096])

The guys with the purple LiSoCl2 batteries make ones with up to 40 years of usable life