洪 民憙 (Hong Minhee)

労働者が機械そのものと資本による機械の使用とを区別することを学ぶには、時間と経験を要した。

——マルクス

150年前のラッダイト運動に対するこの評価が、いまのLLM論争にそのまま当てはまると思っています。「完全な世界で唯物論的に行動すること——生産手段としてのLLMと社会的関係」

勞働者(노働자)들이 機械(기계) 自體(자체)와 機械(기계)의 資本主義的(자본주의적) 適用(적용)을 區分(구분)하는 法(법)을 배우기까지는 時間(시간)과 經驗(경험)이 必要(필요)했다.

—마르크스

150年(년) 前(전) 러다이트 運動(운동)에 對(대)한 이 評價(평가)가 只今(지금) LLM 論爭(논쟁)에 그대로 適用(적용)된다고 생각합니다: 〈不完全(불완전)한 世上(세상)에서 唯物論的(유물론적)으로 行動(행동)하기: 生產(생산) 手段(수단)으로서의 LLM과 社會的(사회적) 關係(관계)〉(한글).

労働者が機械そのものと資本による機械の使用とを区別することを学ぶには、時間と経験を要した。

——マルクス

150年前のラッダイト運動に対するこの評価が、いまのLLM論争にそのまま当てはまると思っています。「完全な世界で唯物論的に行動すること——生産手段としてのLLMと社会的関係」

勞働者(노働자)들이 機械(기계) 自體(자체)와 機械(기계)의 資本主義的(자본주의적) 適用(적용)을 區分(구분)하는 法(법)을 배우기까지는 時間(시간)과 經驗(경험)이 必要(필요)했다.

—마르크스

150年(년) 前(전) 러다이트 運動(운동)에 對(대)한 이 評價(평가)가 只今(지금) LLM 論爭(논쟁)에 그대로 適用(적용)된다고 생각합니다: 〈不完全(불완전)한 世上(세상)에서 唯物論的(유물론적)으로 行動(행동)하기: 生產(생산) 手段(수단)으로서의 LLM과 社會的(사회적) 關係(관계)〉(한글).

It took time and experience before the workers learnt to distinguish between machinery and its employment by capital.

—Karl Marx

This verdict on the Luddites, written 150 years ago, maps onto the LLM debate with uncomfortable precision: Acting materialistically in an imperfect world: LLMs as means of production and social relations.

Random question, but I wondered:

Do you use CAPSLOCK on a physical keyboard?

I have just completed the "Learning the Basics" and "Creating a Microblog" tutorials on @fedify. The Fediverse is very complicated. However, building the example application with Fedify is much simpler, and the tutorial was really good, with lots of examples and explanations of the basics. If you want to check it out, here's the link: https://fedify.dev/tutorial/microblog.

Thank you for creating it, and please consider following @hongminhee!

#fediverse

@liaizon If you have a GitHub account you can report here!

今年中に日本で出版予定のFedifyの本の表紙イラストのラフ案をもらったんだけど、どれもすごく可愛くて選ぶのが大変！

In fact, Hackers' Pub is a bleeding-edge testbed for Fedify. 🙄

Bookmarked: [brandon.si] Linking Smaller Haskell Binaries

https://brandon.si/code/linking-smaller-haskell-binaries/

See more links at https://abhinavsarkar.net/linkblog #linkblog #linkblogging

@z9mb1 Congrats! 🎊

Internationalise The Fediverse

We live in the future now. It is OK to use Unicode everywhere.

It seems bizarre to me that modern Internet services sometimes "forget" that there's a world outside the Anglosphere. Some people have the temerity to speak foreign languages! And some of those languages have accents on their letters!! Even worse, some don't use English letters at all!!!

A decade ago, I was miffed that GitHub only supported some ASCII characters in its project names. There's no technical reason why your repo can't be called "ഹലോ വേൾഡ്".

Similarly, I'm frustrated that Mastodon (the largest ActivityPub service) doesn't allow Unicode usernames and has resisted efforts to change.

So I built a small ActivityPub server which publishes content from an Actor called @你好@i18n.viii.fi - it is only a demo account, but it works!

Some ActivityPub clients report that they are able to follow it and receive messages from it. Others - like Mastodon - simply can't see anything from it. Take a look at the replies on Mastodon to see which services work. You can also see some of its posts on the Fediverse.

What Does The Fox Spec Say?

The ActivityPub specification says:

Building an international base of users is important in a federated network. Internationalization

I can't find anything in the specifications which limits what languages a username can be written in. But there are a few clues scattered about.

The user's @ name is defined by preferredUsername which is:

A short username which may be used to refer to the actor, with no uniqueness guarantees. 4.1 Actor objects

There's nothing in there about what scripts it can contain. However, later on, the spec says:

Properties containing natural language values, such as name, preferredUsername, or summary, make use of natural language support defined in ActivityStreams. 4. Actors

So it is expected that a preferred username could be written in multiple scripts. Which implies that the default need not be limited to A-Z0-9.

The ActivityStreams specification talks about language mapping.

Finally, the ActivityPub specification has some examples on non-Latin text in names.

So, I think that it is acceptable for usernames to be written in a variety of non-Latin scripts.

But What About...?

There are usually a few objections to "Unicode Everywhere" zealots like me. I'd like to forestall any arguments.

What about homograph attacks?

Well, what about them? ASCII has plenty of similar looking characters. I doubt most people would notice when a capital i is replaced by a lower L - and vice-versa. Similarly the kerning issue of an r and n looking like an m is well known. Are mixed language homographs more dangerous? I don't think so.

What if people make names that can't be typed?

Well, what if they do? Maybe not being found by people who can't type your language is a feature, not a bug. But, anyway, clients can let users search for other people, or copy and paste their names.

What about weird "Zalgo" text?

It is up to a client to decide how they want to render text input. The "problems" of strange Unicode combinations are well known. This is not a hard computer-science problem.

What about bi-directional text?

The spec makes clear this is allowed.

Do people even want a username in their own script?

I have no evidence for this. But I bet you'd get pretty frustrated if you had to switch keyboard just to type your own name, wouldn't you? In any case, why can't I have a username of @😉

What's Next?

If you build ActivityPub software, give some thought to the billions of people who don't have names which easily fit into ASCII.

If your software can see @你好@i18n.viii.fi and its posts, please let me know.

@mcc @Hearth @xgranade I'm guessing Q and R are disallowed to mitigate homoglyph attacks. Maybe Old Hangul too, which presumably contains some homoglyphs with modern Hangul.

Somebody linked me RFC 7565, which linked to RFC7564, and if that's the place to look this appears to be the list of disallowed characters in a Fediverse username, and I'm cracking up because it's *mostly* stuff you'd expect, except the very first category of banned characters, specially, is "pre-1700 Korean characters".

The fediverse is welcome to all. EXCEPT KOREAN TIME TRAVELERS. Did you just wake up from being frozen in ice during the Joseon dynasty? The IETF is targeting you PERSONALLY

I Guess I’m now the first ever Indiekit Instance on the #fediverse

Thanks to Fedify

{
"@context": [
"https://www.w3.org/ns/activitystreams",
"https://w3id.org/security/v1"
],
"type": "Person",
"id": "https://rmendes.net/",
"preferredUsername": "rick",
"name": "Ricardo Mendes",
"url": "https://rmendes.net/",
"inbox": "https://rmendes.net/activitypub/inbox",
"outbox": "https://rmendes.net/activitypub/outbox",
"followers": "…

https://rmendes.net/notes/2026/02/19/8b1d5

@evan @cwebber @kopper @hongminhee Couldn’t we agree to standardize on expanded json-ld? We would not need any json-ld processor, we would not need to fetch or cache any context. There would be no way to shadow properties.

@kopper @evan @cwebber Ah, yes, the https://w3id.org/identity/v1 context... As I recall, I probably used it because I needed some kind of context to handle Linked Data Signatures. It's a bit embarrassing, but Fedify also preloads and uses that context. Personally, I wish https://w3id.org/identity/v1 would redirect to a different URL instead of the now-defunct https://web-payments.org/contexts/identity-v1.jsonld.

Nice that @kagihq Translate is available on iOS now. Using it as my new default.

https://apps.apple.com/se/app/kagi-translate/id6748310237?l=en-GB

@kopper @hongminhee As the person probably most responsible for making sure json-ld stayed in the spec (two reasons: because it was the only extensibility answer we had, and because we were trying hard to retain interoperability with the linked data people, which ultimately did not matter), I agree with you. I do ultimately regret not having a simpler solution than json-ld, especially because it greatly hurt our ability to sign messages, which has considerable effect on the ecosystem.

Mea culpa :\

I do think it's fixable. I'd be interested in joining a conversation about how to fix it.

One of the most interesting areas of exploration in the ActivityPub community right now is the ActivityPub API. Most people who know ActivityPub are familiar with its federation protocol, which lets social networking servers like Mastodon and Pixelfed share data between them. But there is another, closely-related feature in the same specification, called the "social API".

ActivityPub has five normative sections: 3 Objects, 4 Actors, 5 Collections, 6 Client-to-Server Interactions, and 7 Server-to-Server Interactions. 3, 4, and 5 provide a read-only interface to social data that is useful for both the federation protocol and the social API. It lets both clients and servers read information about users on the network, their feeds, and the things they make and share.

Section 6 is focused on the mechanism clients can use to create new activities, and the side-effects of those activities. "Activities" are the most important data structure in ActivityPub (which is why they're featured so prominently in the name!). They represent sentences or statements about things that happen on a social network, like "Christine created the image img123.jpg" or "Evan liked Christine's image img123.jpg" and "Amy shared Christine's image img123.jpg". Creating these statements is how clients can make things happen with ActivityPub.

Section 7 is focused on how and when servers can send these activities across the network to other servers. There are some side-effects that are laid out, but mostly they involve cache management.

So, here's the important point I want to make: the federation protocol which connects ActivityPub servers is defined in sections 3, 4, 5, and 7. The social API is defined in sections 3, 4, 5, and 6. But some people use "server-to-server" or "s2s" as a synonym for the federation protocol, even though "server to server interactions" only covers one section. Similarly, some people use "c2s" or "client-to-server" as a synonym for the social API, even if "client-to-server interactions" is only one section.

I prefer to use "social API" or "ActivityPub API" to refer to the entire part of ActivityPub that lets client apps talk to social servers. Here are some rough reasons why.

• It's what the spec calls it. In the conformance section, we said that 'this specification defines two closely related and interacting protocols: A client to server protocol, or "Social API" [...] A server to server protocol, or "Federation Protocol"'. This was intentional; the names come from the Social Web Working Group charter deliverables. Making a Social API was a key goal of the group. The ActivityPub API satisfies that goal.
• It's a term that all developers and many users are familiar with. People who've used mobile apps or third-party apps know what an API is and what it's for. If you use "c2s", it's not clear to most developers what you're even talking about.
• It's accurate. The social API is a RESTful API that uses JSON, HTTP, and all the fun stuff that app developers are already familiar with.
• It emphasizes what's already available. People often say that Mastodon does not implement the ActivityPub API. But it implements sections 3, 4, and 5 of the specification – it has to, in order to support the federation protocol. That's the entire read-only side of the API. All (almost all?) Fediverse servers support these parts of the spec, too. You can build a pretty OK read-only API client using what's already available from Mastodon and others; see https://acct.swf.pub/ for an example.
• "c2s" doesn't cover the whole API. As I said above, only one section of the doc is called "Client-to-Server Interactions", and it doesn't cover the read-only side of the API.
• "c2s" is insider jargon. c2s isn't a common term for RESTful APIs. It really isn't a common term at all; there are only a few protocol suites, like XMPP, that refer to the "c2s" and "s2s" part.

I think it's fine if others use "c2s" when talking about the API, or especially about section 6 of the ActivityPub spec. It's not going to cause any harm. But the Social Web Community Group task force on implementing the API is called the "ActivityPub API" task force. I think that's a good idea – it emphasizes the API. I intent to use this name and framing for the foreseeable future.

Speaking of the quality of Apple's Private Cloud Compute model…

🫥

Discordはもともとクローズドなので、開かれたコミュニティ志向でやってるプロジェクトで採用するのはどうかって話ではあったかも。

コミュニティをDiscordからMatrixへ段階的に移行しています。メンテナーとコントリビューターはすでにMatrixに移っているため、今後はMatrixのほうが返答が早くなります。Discordはしばらく継続しますが、Matrixがメインの場となりました。

詳細とMatrixルームの一覧はこちら：https://github.com/fedify-dev/fedify/discussions/573（英文）

저희 커뮤니티를 Discord에서 Matrix로 조금씩 이전하고 있습니다. 메인테이너와 기여자들은 이미 Matrix로 옮긴 상태라, 앞으로는 Matrix 쪽이 응답이 더 빠를 거예요. Discord는 당분간 유지되지만, Matrix가 이제 메인 거점입니다.

자세한 내용과 Matrix 룸 목록은 여기서 확인하세요: https://github.com/fedify-dev/fedify/discussions/573.

We're gradually moving our community from Discord to Matrix. The maintainers and contributors are already there, so you'll get faster responses on Matrix going forward. Discord will stay up for a while, but Matrix is now our primary home.

For the full details and the list of Matrix rooms, see: https://github.com/fedify-dev/fedify/discussions/573.

What if Mastodon API Enhancement Proposals so that we won’t have 5+ different ways to change your password on platforms implementing MastoAPI?

@FredPraca 언젠가 韓國(한국)에 꼭 놀러오세요! 😀

@FredPraca 炸醬麵(짜장면)은 韓國式(한국식) 中華料理(중화요리)라서, 韓國(한국) 밖에서는 찾기 어려울 거예요. 😅