Daphne Preston-Kendal

a human pretending to be a kitten, which in turn is pretending to be a human.
or maybe the other way around

this account cannot decide if it is professional or personal.
it is definitely unprofessional and may be impersonal

Computers are bad and you shouldn’t use them
but if you do, you should definitely use Codeberg (presidium member 2025–27)
and hack in Scheme (once and future WG2 chair)

🇪🇺 (🇩🇪←🇬🇧) ✝️ 🌹 ☂️ ↙↙↙ 🇵🇸🇺🇦 🏳️‍⚧️🐱

Location: Berlin, Germany
Pronouns: she/sie/elle/ji/hēo/hon/ἥ/सा

Public Key

npub1ly9acvqg8rwnanesz6cvxyhpk2ndqprwpk5lfs8d2s8xcygnkqyqqkktws

NIP-05 Address

dpk_at_chaos.social@momostr.pink

Profile Code

nprofile1qqs0jz7uxqyr3hf7eucpdvxrztsm9fksq3hqm205crk4grnvzyfmqzqpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0ds67keqg

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1ly9acvqg8rwnanesz6cvxyhpk2ndqprwpk5lfs8d2s8xcygnkqyqqkktws

Show more details

Last Notes

2026-04-02 19:39:49 UTC

- reply

They made a declarative grammar, and the language it describes is regular, and the spec doesn’t assume a particular application. That’s more than WHATWG has ever managed

2026-04-02 18:36:58 UTC

- reply

Ah, crud, unfortunately there is a part of URL resolution which will break both . and .. https://datatracker.ietf.org/doc/html/rfc3986#section-5.2.4

2026-04-02 16:21:30 UTC

How much will people hate me if I make /./ a significant component of Hasgit’s URL design

2026-04-01 10:37:21 UTC

- reply

Isn’t it curious how Apple has brought out its cheapest laptop ever at a time when hardware is otherwise at its most expensive in a long time? It’s an underpowered client for cloud services
Welcome to the future of computing. The ‘thin client’ dream of the 90s is here!

2026-04-01 10:35:08 UTC

- reply

Cloud providers will have identified self-hosting, community-organized services, and digital sovereignty campaigns as a threat, in the same way Microsoft identified free software as a threat in the 1990s and 2000s, and they will want to stamp out that threat. Keeping prices high to make self-hosting uneconomical is one way to do this

2026-04-01 10:32:29 UTC

- reply

The cloud providers and the AI companies are the same companies, and what they have learned is that they can tie up all the high-end hardware for themselves and make it unaffordable for plebs
The hardware companies in turn will be happy to keep prices high even if raw materials are ever more easily available again and demand is (somewhat) lower

2026-04-01 10:28:16 UTC

The hardware shortages and prize hikes are not going to end with the AI bubble

2026-03-31 06:12:12 UTC

Anyone have experience with employee buyouts of small businesses in Germany by any chance

2026-03-27 18:56:49 UTC

- reply

I don’t foresee any problems there, but I‘m not deep in the technical issues of federation. Perhaps @npub1yru…hgt9 could say more, although I suspect that sort of thing is still a way down the development roadmap

2026-03-27 18:17:11 UTC

- reply

(This thread reflects only my own views and not those of the remainder of the Codeberg presidium, board – nor (most importantly) does it reflect the views of the members of Codeberg as a whole, who ultimately make the decisions.)

2026-03-27 18:17:04 UTC

- reply

So no, we don’t offer massive CI resources, unlimited repo quotas, etc.
What we offer is different: if self-hosting isn’t for you, but you still want to know that your repo hosting is not at the whims of Microsoft’s bottom line, Codeberg is for you.
If the Codeberg service you depend on goes away, it won’t be from one day to the next in order to boost our income. We will warn you about the reasons, ask you and the other members what you think, and you’ll have a chance to change things.

2026-03-27 18:13:18 UTC

- reply

Codeberg is different.
Codeberg is a non-profit.
Moreover, it is a *democratic* non-profit. You can (and, if you actively use Codeberg, should) become a member and you will have an equal vote and a fair say in the running of the association, and in any decisions we make about what to offer and what not to offer.
This democratic right is guaranteed to you not only by our constitution, but by German law, which guarantees that the membership as a whole has the ultimate decision-making power.

2026-03-27 18:10:12 UTC

- reply

Microsoft’s shareholders are probably very pleased about this.
But if Microsoft needs to please its shareholders and turn the revenue dial up a bit again (say, because they’ve got themselves tied up in a huge economic bubble, and need to bolster the bottom line after it pops), they might choose to take away some of those free goodies.
In fact, it’s not just a hypothetical – they started doing that just recently. https://web.archive.org/web/20260223084443/https://github.com/resources/insights/2026-pricing-changes-for-github-actions
Those changes are ‘postponed’, but assume they will come.

2026-03-27 18:05:04 UTC

- reply

Microsoft is a for-profit company. They have attracted a lot of projects to them by giving away free goodies, like an immense amount of computing power for CI, and over the years have increased the amount of stuff you get for free. (Not long ago you had to pay for personal private repos …)
By continuously sweetening the deal for smaller users, then getting orgs to pay once they start needing collaboration features, GitHub built up over a billion dollars in revenue per year (as of 2022).

2026-03-27 17:55:22 UTC

- reply

Apart from that, many people listed things that GitHub provides for free which Codeberg doesn’t provide or requires self-hosting for. The biggest one was CI, but there were other things too.
And I think the only way I can respond to that is by talking about the fundamentally different organizational models that GitHub and Codeberg have – not to plead on our smaller size, but to plead on principles.

2026-03-27 17:46:48 UTC

- reply

(If someone’s forge were hosting completely objectionable, hateful content we might choose to turn their OAuth support off. But, for example, you can have as many private and non-free repositories as you like that way. More on that in a bit though …)
(Furthermore, this thread reflects only my own views and not those of the remainder of the Codeberg presidium, board – nor (most importantly) does it reflect the views of the members of Codeberg as a whole, who ultimately make the decisions.)

2026-03-27 17:45:48 UTC

- reply

Forge federation is something we’re actively pushing for because of this. Full federation support is still a few years ahead of us, probably, but you can already set up your own forge in a way that minimizes inconvenience to users by using Codeberg as a single sign-on provider for a self-hosted forge.
https://git.madhouse-project.org/, which hosts the Iocaine project, is a great example of this. If you set up Codeberg as SSO for your own site, almost all of our content rules are irrelevant to you.

2026-03-27 17:44:06 UTC

- reply

We are much more excited by the idea of creating an ecosystem of forges (and different forge software) than in becoming the new one place where absolutely everything is hosted.
I am, for example, very pleased that even while switching to Forgejo, Fedora decided to keep on self-hosting rather than jumping to us.
To be clear, we’re open for everyone who needs us, especially individuals working on their own smaller projects. But other forges are our friends, not our competition

2026-03-27 17:38:47 UTC

- reply

To begin by addressing what seemed to be a misconception underlying a lot of the comments:
Codeberg, the website, does not aim to replace GitHub. We poke fun at their recent foibles in our PR sometimes, but we do not want to become them.
We do not want to replace one single point of failure with another! (Admittedly, I think Codeberg.org, run by a non-profit, as SPOF would be better than GitHub.com, run by Microsoft, as an SPOF, but that’s not the goal!)

2026-03-27 17:36:02 UTC

- reply

2026-03-27 17:34:56 UTC

There was an article about moving to Codeberg on the orange website yesterday, and out of interest to get an impression of what average programmers think of the idea of moving, I took at look at the comments. (Yeah, I know …)
I’m going to write here my thoughts on a general impression of the comment thread as a whole. (Rather than respond to individuals.)

2026-03-06 16:49:45 UTC

- reply

‘Herr’ in German doesn’t have the same effect, although it’s admittedly basically calling God ‘Mister’

2026-03-06 16:48:50 UTC

I find Christian prayer in French somehow unseemly (e.g. ‘Seigneur’ as an address for God just sounds unaccountably silly to me), but Duruflé’s short setting of the Lord’s Prayer goes incredibly hard

2026-03-06 11:46:57 UTC

Kernel security bugs are fixed three times faster when fixed by the same person who wrote the bug https://pebblebed.com/blog/kernel-bugs-part2
Applying this result to a world in which devs aren’t the authors of ‘their’ own (slop) code left as an exercise to the reader

2026-02-18 12:17:34 UTC

- reply

What do you do when describing the timeline updates of users with ‘they/them’ pronouns to users who have their language set to a language without a gender-neutral personal pronoun?

2026-02-18 12:17:06 UTC

Another thorny accessibility-adjacent problem:
You have a web app like a social network where there’s a timeline that describes recent things people have done. ‘Alice updated her profile’, etc.
Users can choose the pronouns the site uses for them. ‘They/them’ is one of the options available.
But you want to make the site available in multiple languages. Some languages only have two gendered personal pronouns.

2026-02-03 09:04:37 UTC

This evening, Andreas Shimokawa and I are talking to the Berlin Permacomputing meetup about Codeberg’s history and future.
19:00, /rosa (Heidelberger Straße 28, 12059 Berlin)
Come and join us!
(Boost appreciated!)

2026-01-04 19:59:02 UTC

Help out your local FOSS nonprofit:
Smash a fucking piledriver into the skull of a nearby AI entrepreneur

2026-01-01 13:00:09 UTC

- reply

It’s a semiprime
Will this do

2025-12-29 16:58:38 UTC

Reminder to everyone on the TL for 2026: It’s going to be another tough year. Don’t forget to have some fun, and don’t feel guilty for it. You need it to survive.
https://assets.chaos.social/media_attachments/files/115/803/974/176/876/600/original/1681828556d742fd.png

2025-12-16 11:06:48 UTC

https://jyn.dev/i-m-just-having-fun/

2025-12-12 11:48:50 UTC

Still a great idea to hold conferences in the US. Furthermore, nobody could have predicted this 12 or 18 months ago, so it’s entirely reasonable to continue with plans that were made in 2024
https://assets.chaos.social/media_attachments/files/115/706/437/772/974/970/original/a7d6f2161db626f3.jpeg

2025-11-21 18:27:41 UTC

Fucking. Hell. https://github.com/ocaml/ocaml/pull/14369

2025-11-02 18:16:19 UTC

- reply

I can recreate the Darwin behaviour on Linux (NixOS) by setting LANG=en_GB.UTF-8, but not the opposite

2025-11-02 18:12:29 UTC

- reply

What’s your $LANG set to on those?

2025-11-01 08:33:34 UTC

- reply

He also thought really hard about how to ask the right question about sex/gender on his 2010 colour survey (because, unlike literally everyone else who thinks your biology is any of their business, he was actually interested in something to do with the Y chromosome), at a time when most people thought trans people didn’t really exist.
https://blog.xkcd.com/2010/05/06/sex-and-gender/

2025-11-01 08:28:30 UTC

I think Randall Munroe deserves more credit than he gets for making his comic, a touchstone of internet nerd culture, strongly against the culture of nerd sexism from its earliest days https://xkcd.com/322/

2025-10-28 09:29:48 UTC

Alan Kay’s critique of SimCity games
https://assets.chaos.social/media_attachments/files/115/451/087/135/746/277/original/6d9a460410daf86f.png

2025-09-27 12:17:01 UTC

- reply

An early unit of computational power was the ‘kilogirl’. (Pictured: 0.006 kilogirls)
https://assets.chaos.social/media_attachments/files/115/276/213/022/296/245/original/9bae43026c5458a2.jpeg

2025-09-27 11:41:20 UTC

- reply

Margaret Hamilton is 89 years old

2025-09-27 11:40:39 UTC

Some things I have learned about women in computing history today:
There is a *photograph* of the Countess of Lovelace
https://assets.chaos.social/media_attachments/files/115/276/069/991/953/036/original/5f863c1713f75856.jpeg

2025-09-25 14:21:33 UTC

Whenever I see another round of the C++ community’s war against guaranteed memory safety, I think of Tony Hoare’s Turing Award lecture where he said that if computer engineering were real engineering, memory safety bugs would be considered criminal negligence.
He did not mean this as hyperbole. If you ship software and it fails in the real world due to a memory safety bug, they should find out who put their engineer stamp on the software and prosecute them criminally.

2025-09-01 18:06:38 UTC

- reply

Alas, no: there are only nine pre-programmed ‘region’ settings (four of which are for specific cities in Japan, and the rest of the world has to share the other five)
https://assets.chaos.social/media_attachments/files/115/130/367/681/515/968/original/9feb1d1070f8a0a3.jpeg

2025-09-01 08:50:21 UTC

- reply

Going to investigate later if the range of options is large enough that e.g. one could use it for cooking on Mars or the Moon

2025-09-01 08:27:17 UTC

- reply

It may not surprise you to learn that the default is set to Japan

2025-09-01 08:24:50 UTC

Fiancé has discovered that his very fancy kitchen scale has a programmable option to set the value of g for your location on Earth

2025-08-12 11:16:46 UTC

- reply

Storing in each entry the key, value, and the hash value of the key, rather than just the key and value. If you don’t store the hash value, you have to call the hash function again on every key when the hash table is resized

2025-08-12 11:09:22 UTC

Results from an initial survey:
Every non-Scheme hash table implementation I’ve looked at (Python, Ruby, OpenJDK) caches the hash value for each entry.
Every Scheme hash table implementation I’ve looked at (Chez, Chibi, Guile) doesn’t cache the hash value for any entry.
🤔

2025-07-31 09:16:11 UTC

Helsinki goes a full year without a traffic death
A city traffic engineer credits the success to lower speed limits and smarter design.
cc @npub1z7x…q5jf
https://yle.fi/a/74-20174831

2025-07-05 19:51:15 UTC

- reply

What did they want? I’m guessing this wasn’t a philanthropic pitch

2025-07-03 13:06:18 UTC

- reply

But as long as conferences are held in person, holding them in countries like the US, Russia, PRC etc. is particularly irresponsible

2025-07-03 13:02:46 UTC

- reply

FWIW, my personal view is that we should also be reducing the number and significance of in-person conferences in general.
Not only because conferences are inherently exclusionary – even if held in a country with an ideal and egalitarian visa policy (no such country exists), travel and accommodation costs restrict participation – but also because of the massive environmental cost of so many people travelling to the same place to see papers they could have read or listened to online

2025-07-03 10:26:58 UTC

ULT.
Do not host international conferences in countries with totalitarian regimes.
Do not host international conferences in countries with totalitarian regimes.
Do not host international conferences in countries with totalitarian regimes.
(And if you do, don’t come later with crocodile tears about ‘academic freedom’.)

2025-04-18 17:16:29 UTC

- reply

Actually, it’s possible that someone else with admin access saw this toot before me and fixed it themselves

2025-04-18 17:13:37 UTC

- reply

r6rs.org and r7rs.org

2025-04-18 17:07:59 UTC

- reply

Okay, now I’m very confused because I checked and those are the details I used for domains I set up earlier this year

2025-04-18 14:35:03 UTC

- reply

Did you mean 19 April 2024 or some other year? 19 April 2025 is tomorrow 🤔

2025-04-16 12:26:28 UTC

- reply

Was there ever any year where web design was better than it was in 1999? I mean just look at this stuff. Unbeaten tbh https://www.webdesignmuseum.org/all-websites/year-1999

2025-04-16 11:21:32 UTC

- reply

Some have asked whether an ECtHR appeal could have any effect. This particular case could probably not be appealed – the Scottish government hasn’t had its human rights violated; it would need a trans person affected by this ruling to make the case, aiui – but actually, on reflection, I think maybe yes. This decision, taken in the full picture of the current legal situation for trans people in the UK, completely shits on the resolution of Goodwin & I. v. United Kingdom

2025-04-16 11:16:59 UTC

- reply

The Gender Recognition Act 2004 is a dogshit piece of law that the New Labour government was forced to introduce by a European Court of Human Rights ruling. It is well in the tradition of UK governments doing the absolute minimum to comply with pan-European law: it was always meant to do the absolute bare minimum for trans people.
Now it doesn’t even seem to do that any more.

2025-04-16 11:14:56 UTC

- reply

So, a real question: What is a Gender Recognition Certificate actually good for now?
You can change your identity documents to have the correct name and gender without one.
It doesn’t make you the correct gender (at least if you’re a woman) for the purposes of positive discrimination.
It seems to be completely irrelevant to protection against negative discrimination whether you have one or not.
Since the Marriage (Same Sex Couples) Act 2013, you don’t need one to marry, either.

2025-04-16 11:11:01 UTC

- reply

So it’s also a not entirely unreasonable reading of this section of the decision to suggest that passing trans people who do not reveal their transgender status can enjoy protection from negative discrimination on the basis of being perceived to be their correct gender – but an openly transgender person, discriminated against because of their transgender status, would have less protection in law.
This section of this particular ruling is unclear to me. It likely depends on a lot of other things

2025-04-16 11:08:13 UTC

- reply

For this protection, a Gender Recognition Certificate is irrelevant, since it is based on perceptions, not facts.

2025-04-16 11:06:24 UTC

- reply

The court’s basis for affirming this protection was existing precedent that negative discrimination is still illegal if based on a characteristic that the discriminating party perceived the discriminated party to have, whether or not that perception is correct. (Concrete [invented] example: if someone were discriminated against on the basis of a Jewish-sounding or Muslim-sounding name, but were not actually Jewish or Muslim, they could still claim religious and racial discrimination.)

2025-04-16 11:02:58 UTC

- reply

Also, the court affirmed certain protections against negative discrimination do continue to apply to trans women. There are two ways this could be read:
The first and most obvious is that trans people of gender x cannot benefit *in general* from positive discrimination in favour of gender x, but can benefit from protection against negative discrimination against gender x.
This would be the better of the two interpretations. Unfortunately …

2025-04-16 11:00:42 UTC

- reply

This case concerned positive discrimination and whether a trans woman appointed to a position with a gender quota would count as a woman. The court said no.
It might be taken to imply that a trans man appointed to a similar position also would not count as a man, but much of the reasoning in the written decision considers only trans women. Maybe a trans man would count as a man if the cause were fought again: stranger things have happened in British law.

2025-04-16 10:54:50 UTC

- reply

The good news:
This case was fought by the Scottish government, which could now introduce primary legislation in Holyrood to reverse the effect of the decision in Scotland. I expect there would be another fight over this: ‘Equal opportunities’ is, with some exceptions, a Reserved Matter in the Scotland Act 1998. They may well be able to get this in under an exception.
Trans people in England and Wales and NI could not benefit from this, though.
We’ll see how the Scottish govt plays this now.

2025-04-16 10:53:03 UTC

- reply

I think both of these are wrong individually – in using the long title of the EA 2010 to ascertain its purpose, it appears to have ignored the potential effect of the word ‘reform’ in ‘harmonise and reform’ – but this is now the law.

2025-04-16 10:46:52 UTC

The two fundamental planks of today’s decision are:
1. the Gender Recognition Act 2004 did not change the meaning of ‘man’ and ‘woman’ in the Sex Discrimination Act 1975, although the relevant section of the GRA on its face did not contain any limitation on the application of a certificate in determining who is a ‘man’ and who a ‘woman’
2. the Equality Act 2010 inherited from the SDA 1975 this definition wholesale, since its purpose was to consolidate existing law

2025-04-15 09:47:39 UTC

- reply

https://assets.chaos.social/media_attachments/files/114/341/344/435/715/183/original/2264543d9a2fac7b.png

2025-02-05 07:58:28 UTC

Is there a name for the inverse of set intersection? I.e. the set of all elements that are in at least one, but not all, of the input sets?

2025-01-25 09:19:54 UTC

- reply

> Camera sensors THIS year might be down grades from camera sensors LAST year, because the chips got more expensive.
Not content with stealing the actual art of real artists, now AI is making the tools of real art worse as well!

2024-12-30 14:54:18 UTC

Okay, once again I am considering getting myself a Mac SE/30.
Question for enthusiasts of comparably-vintage Macs: How do you transfer files to/from a modern Mac? Modern Mac OS supports neither the original HFS, nor AppleTalk (if I could even get them hooked up for that). Am I stuck with System 7+ and using FAT/PC Exchange (presumably with StuffIt files or something to preserve the resource fork)?

2024-12-23 00:14:22 UTC

- reply

Please take another look at Ed Davies’s comment. The problem is codepoints being represented in more bytes than the minimum for the encoding of that codepoint. The letter ‘a’ MUST only be representable by the single byte x61. The two bytes xC1 xA1 will decode to the same codepoint in a naïve decder, but MUST be rejected by a conforming decoder. This happens *within* the 2 to 4 byte long range of 20.whatever-bit UTF-8.

2024-12-22 21:48:18 UTC

- reply

I can’t comment on your blog for some reason, so I’ll answer here:
Normalization and normal forms has nothing to do with overlong UTF-8 sequences. Normalization has to do with valid code unit sequences which represent different codepoint sequences which in turn represent different ways of writing a character which humans generally consider ‘the same’.
The example Ed Davies gave is invalid UTF-8: a conforming decoder should reject it, for reasons touched on in section 10 of RFC 3629.

2024-12-13 17:53:35 UTC

- reply

Was RFC 9485 your attempt to ensure regular expressions remain xkcd 927-compliant?

2024-10-14 17:36:07 UTC

I fully endorse and support the recommendations of the Stallman Report (https://stallman-report.org/, [@report_press](https://mastodon.social/@report_press )).
I also once again urge Scheme implementations which are affiliated with the FSF (USA) or the GNU project to disaffiliate.
(This toot is posted in a personal capacity only.)

2024-10-12 21:51:28 UTC

- reply

Why? I am not looking forward to telling everyone in my community ‘oh, we were on Codeberg, but they changed their name and now we’re on [whatever new name you’ve come up with]’

2024-09-25 22:11:14 UTC

- reply

Precedent: the regexp processor of the MOO programming language used (uses!) % instead of \. I think it got this from somewhere else before it, too, but I don’t know where.
Why are you planning to write your own regexp processor instead of using Go’s? It is good and linear time (although the constant factors aren’t great in comparison to some others). You can even construct an AST from your own regexp syntax and let it do the NFA/DFA conversion and simulation for you