Matthew Garrett

https://nondeterministic.computer/@mjg59

Former biologist. Actual PhD in genetics. Security, OS security teaching at https://www.ischool.berkeley.edu. Blog: https://codon.org.uk/~mjg59/blog . He/him. Opinions do not reflect those of my employer.

Blog: https://codon.org.uk/~mjg59/blog
Signal: @mjg.59

Public Key

npub1jpfx7gvwn68l9rw8def53lkljwg0yv7fn99tqdzxyn68k29p8gmq9qvqhj

NIP-05 Address

[email protected]

Profile Code

nprofile1qqsfq5n0yx8farlj3hrku56glm0e8y8jx0yejj4sx3rzfarm9zsn5dspqq5jewhe

Publishing to

Author Public Key

npub1jpfx7gvwn68l9rw8def53lkljwg0yv7fn99tqdzxyn68k29p8gmq9qvqhj

Show more details

Last Notes

2026-07-02 10:32:48 CEST

RE: https://fediverse.zachleat.com/@zachleat/116845291090365543
Obviously everyone knows that Vic Gundotra killed Reader, but he didn't succeed immediately. He could destaff the project, but people could still spend 20% time maintaining it. It was inevitable that it would die, but what the final blow was a standard Google outcome - load bearing infrastructure was going to be turned off, and everyone had to port their project to the replacement that didn't really work yet. Reader just didn't have the resources, so died.
#note1d2d…60ks

2026-05-07 17:56:46 CEST

The downside of full-on self-hosting is, of course, that your server will inevitably choose to have weird hardware issues while you're in another country

2026-04-27 10:12:54 CEST

Speaking of Unicode, I think many people would be surprised just how strong the pushback against it was even in the early to mid 2000s. UTF-8 adoption in Linux was a fight.

2026-04-18 11:27:43 CEST

- reply

Clearly my most unpopular thread ever, so let me add a clarification: submitting LLM generated code you don't understand to an upstream project is absolute bullshit and you should never do that. Having an LLM turn an existing codebase into something that meets your local needs? Do it. The code may be awful, it may break stuff you don't care about, and that's what all my early patches to free software looked like. It's ok to solve your problem locally.

2026-04-18 10:25:15 CEST

- reply

(Yes ok there are cases where code is beauty and embodies an idea that could make a grown man cry and:
(1) your code is not that code
(2) you would think nothing of copying the creative aspect of that code if you needed to don't fucking lie to me)

2026-04-18 10:20:31 CEST

- reply

Look, coders, we are not writers. There's no way to turn "increment this variable" into life changing prose. The creativity exists outside the code. It always has done and it always will do. Let it go.

2026-04-18 10:14:40 CEST

Free software people: A major goal of free software is for individuals to be able to cause software to behave in the way they want it to
LLMs: (enable that)
Free software people: Oh no not like that

2026-04-17 07:09:24 CEST

RE: https://infosec.exchange/@0xabad1dea/116418277310717935
The secret third party paying my legal bills has finally been revealed
#note15xg…7amt

2026-03-16 09:10:56 CET

I am watching an absolutely fascinating exchange between a domain expert and someone driving Claude compete to produce the smallest functional implementation of a well-defined problem and wow the robot is being competitive, but this is very much a best case scenario of there being a very clear functional specification the output can be tested against

2026-02-23 05:58:01 CET

- reply

The nature of side channel attacks is that CPU-imposed barriers are no longer as strict as they should be. This means that hypervisor boundaries are porous - it's possible to extract material from other virtual machines. I understand the perspective that simply avoiding executing any untrusted software removes that risk, but I do not control the software running in other VMs on the same hardware. How do you solve that?

2026-02-21 12:05:12 CET

18 years of the Linux fork that refuses to load the microcode that mitigates a range of CPU vulnerabilities

2026-02-06 04:45:26 CET

- reply

It doesn't matter whether C is good or not. It matters that if I write code in two languages that aren't C, and I want it to all be part of the same process, I need to care about C. C pervades all. You cannot escape it. C will outlive all of us. The language will die and the ABI will persist. The far future will involve students learning about C just to explain their present day. Our robot overlords will use null terminated strings. C will outlive fungi.

2026-02-06 04:34:17 CET

https://faultlore.com/blah/c-isnt-a-language/ deserves a fucking record for managing to trigger people into being extremely upset while also demonstrating that they don't understand the actual point being made

2026-02-03 01:11:56 CET

- reply

Ah! https://gitlab.freedesktop.org/hadess/iio-sensor-proxy/-/issues/216 ok I'll dig into this

2026-02-02 21:48:49 CET

[@hadessuk](https://hachyderm.io/@hadessuk ) I have a tablet that exposes two accelerometers via iio which are supposed to be used to identify whether it's in tablet mode or not. Any idea what a generalisable solution to that would look like? There's various hacky device specific ones.

2026-01-26 05:49:45 CET

- reply

But merely being free software isn't sufficient - software developed in a way that prevents arbitrary observers from witnessing design conversations may still be free software, but doesn't give us a strong reason to trust the developers. We all know how easy it is to hide dubious code in the open. The libxz backdoor was discovered by examining the binary and tracking that back to the source, not through source examination.

2026-01-26 05:46:56 CET

The presumption that free software is sufficient or necessary to ensure all software you depend on is trustworthy is simultaneously naive and ignorant of what software is capable of. The only realistic way to develop trust in software is to trust the people who write it, and development processes associated with free software make that trust easier.

2026-01-15 10:54:04 CET

Can someone please just write an ebpf script that blocks ../ and make it possible to automatically load that in systemd units and docker and any other security boundary

2026-01-13 08:00:13 CET

- reply

So: In the past the infrastructure was simpler, the underlying hardware was simpler, the number of use cases you had to satisfy was smaller. And now everything is fundamentally more complicated and you're competing with platforms that have millions of absolutely normal computer users using them. But we've also got greater avenues of sharing knowledge, collaboration, better understanding of how to build abstractions. Was the golden age 30 years ago, or is it now?

2026-01-13 07:45:34 CET

- reply

And yeah there's all sorts of cases where X exposes functionality that doesn't really exist in Wayland and I understand people wanting that but also I shared my desktop with some people today on a call and got a notification icon letting me know it was happening and I didn't have to rely on the client to behave correctly for that to be something I could rely on

2026-01-13 07:42:59 CET

- reply

I'm pretty fascinated about why this is happening and my gut feeling is that people have simply written better abstraction layers on top of Wayland than were possible on top of X and now people can just write shit without needing to care anywhere near as much about the sharp edges that exist everywhere

2026-01-13 07:41:40 CET

It is actually kind of wild that we're simultaneously in an era of people complaining that Wayland is destroying choice and also maybe the greatest number of high-quality desktop environments aimed at different use cases the free software world has ever had

2025-11-28 06:51:38 CET

Poll for people from the UK and US only sorry I am being restrictive there is actually a reason

2025-07-28 00:20:29 CEST

Need my GB driver's license details For Reasons, can't find the card, remember it's saved in my Hertz profile, go there, page shows me most the the number ***d out, check developer tools, API is returning the full number and just obfuscating it client side, hurrah

2025-07-07 08:57:37 CEST

"Huh why is logging into the server in the room next door so slow" because my eeros rebooted overnight and my laptop roamed onto the other SSID that's VPNed back to the UK and all my packets have been travelling 10,000 miles to go a few metres

2025-06-09 23:03:28 CEST

- reply

no meaningful secure messenger asserts that messages cannot be shared by the recipient, and any developer of such an app that made that claim would be an indication that app shouldn't be trusted

2025-06-09 13:40:50 CEST

- reply

the enclave attests itself to the client, the server admin can't fake it or monitor the communication

2025-06-07 13:57:58 CEST

- reply

yes, you can trade off functionality to gain privacy on a fairly arbitrary scale, but teaching people to make informed decisions about where they should be on that scale is hard and you still end up with metadata transfer between servers with very different security properties. Is this a good choice for some people? Yes! Are most people going to end up making well informed choices? Not even slightly

2025-06-07 12:54:20 CEST

- reply

you're trading off centralised server infrastructure for contact discovery that's less privacy preserving (quicksy has a list of every Jabber ID and phone number combo!), which may be worth it for some use cases but certainly isn't for all!

2025-06-05 13:12:20 CEST

Twitter's new encrypted DM system stores your private key material on Twitter-owned services, protected with nothing more than a 4-digit PIN. If hostile, or if legally compelled to, Twitter could easily decrypt all your messages. It's also MITMable and doesn't secure metadata. Use Signal.
https://mjg59.dreamwidth.org/71646.html

2025-05-24 08:40:21 CEST

- reply

Yeah great ok the manufacturer also has no ability to patch my fridge but how does that help me this differentiation between software in ROM and software in flash is absolute bullshit, it's all software and it should all be free and modifiable

2025-05-19 09:34:52 CEST

Ah good flight to the UK booked for next month

2025-03-21 17:10:30 CET

- reply

If the membership chose a candidate that the board refuses to appoint, the membership deserves to know that.

2025-03-21 17:07:14 CET

- reply

There is a well defined process for how elections are run (https://opensource.org/about/board-of-directors/elections). This year, several candidates were removed from the election results for supposedly refusing to use proprietary software to sign the board agreement. The requirement to do so is not stated anywhere in the election process. The bylaws allow the board to ignore the will of the electorate, but removing candidates who didn't violate any election rules is just fucking ridiculous.

2025-03-21 17:00:25 CET

Like many nonprofits, membership of the @npub1djq…zj5f board of directors is technically down to the existing board choosing to appoint new directors. OSI "membership" (in the sense that donors can become "members" of the organisation) isn't a thing that's defined by the by-laws, it's a process that the directors have chosen to adopt. As such, the elections that members participate in are just guidance for the board's decisions, rather than anything they're obliged to respect.

2025-03-19 01:01:10 CET

When Twitter launched encrypted DMs they were bad. They haven't improved. The person behind them is now a senior member of DOGE and getting appointed to the board of a government-backed mortgage giant: https://mjg59.dreamwidth.org/71188.html

2025-03-18 04:05:47 CET

I'm a huge proponent of copyleft licensing, I'm in favour of using the GPL as a tool to ensure users have the ability to modify the software on their devices, and I'm just having trouble getting too worked up about the Rust reimplementation of Coreutils being MIT. Philosophically? Yeah, it sucks. Practical outcomes? Almost certainly none. The GPL violators aren't going to change coreutils implementation to avoid being sued, the FSF wasn't going to do that anyway

2025-03-08 19:46:45 CET

The discovery that lifepo4 power banks are now at the point where they can switch from AC fast enough that they're usable as UPSes is about to change my life in a short term expensive but long term cheap way

2025-03-02 06:11:48 CET

Free software is about granting people rights, and as such it is incompatible with racism. It is incompatible with homophobia. It is incompatible with "gender critical" ideology. Those people all reject the idea that all humans have rights based on who they are are ideologically incompatible with free software and deserve to be sidelined. We can replace anyone's contributions, we can't replace all those they drive away.

2024-08-31 05:35:05 CEST

Rust advocates should demonstrate the viability of the language by replacing an entire kernel. One that meets accepted architectural goals. One that is at the core of the free software movement.
Rewrite Hurd in Rust, you cowards.