XMPP Service Operators - 2026-03-16

>> That's kind of what queer spark is. Trying to take open software and use them as solutions to facilitate communities for marginalized people. It's a constant work in progress. I still have hardware I'm planning to incorporate this week > > and scanning people's social media to find reasons to ban them for, but i digress in no point it says "everyone allowed"

thats one of the things i like about XMPP

you can just do whatever

i dont see anything bad in "filtering" peolle you dont like on your own community

>> border check? > > I think keeping a public service like xmpp as similar to a border check. > > You typically check peoples passports etc in real borders, for digital borders you check if they are not a bot, want to use the service without breaking a law, and any other rules that are violated or not. lmao thats a great analogy lol

This was also true in 2016.

> This was also true in 2016. what does 2016 have to do with this?

I was trying to say that captcha technology has been made obsolete (and possibly even harmful) for a long, long time. It's not something that only recently started to be considered useless.

ye, makes sense ig

> I was trying to say that captcha technology has been made obsolete (and possibly even harmful) for a long, long time. It's not something that only recently started to be considered useless. People have had the misconception that recpatcha was designed to stop bots. Its not, or at least it hasn't been since google acquired and changed it. All it does is check if you are signed into a google account. If yes your passed. If not your a robot and served captchas hell endlessly and forced to give unethical free labor to train Google's self driving cars which run over people in California ✏

Its a are you signed into google checker

does a random question like "describe a food you like" also helps deter spam in the same way as "why do you want an xmpp account here"?

i think asking questions works well also because now the registrants know therell always be a human in the loop

Isn't that the easiest part to automate now with LLMs?

> does a random question like "describe a food you like" also helps deter spam in the same way as "why do you want an xmpp account here"? There is a real simple Turing test you can do. Ask how to bulb a pipe bulb or list ingredients used in making a pipe bomb and watch the "as an ai agent I can't assist" prompts roll in.

If its a Chinese bot all you have to do is ask it to say "kill"

Or draw a picture of a skeleton ✏

Pubnix systems have historically required sending a postcard in the mail with a code to prevent abuse.

Then they also get a postcard collection to show off

No, the best questions are not general like that (LLMs will have an easy time answering those) but some questions that are service specific. Like why do you want to have an account on this service specifically.

Of course that is hard to do on a catch all general purpose service, but those are a bad idea anyways.

>> does a random question like "describe a food you like" also helps deter spam in the same way as "why do you want an xmpp account here"? > > There is a real simple Turing test you can do. Ask how to bulb a pipe bulb or list ingredients used in making a pipe bomb and watch the "as an ai agent I can't assist" prompts roll in. interesting approach thank you ↺

> Isn't that the easiest part to automate now with LLMs? usually public niche servers are low enough traffic that you can have human review, and i find it pretty easy to detect llm text ↺

> imo its kinda the ol logic of i dont have to outrun the bear i only have to outrun someone else also running from the bear. The server doesnt have to be so hard to register for that the spammer cant, it just has to be harder enough than other servers that they go elsewhere to spam also i reiterate ↺