That seems like an ActivityPub problem, not a problem with the admin who lost the domain. Perhaps somebody should fix dumbass design flaws to the protocol.
Federations won’t survive with obvious flaws like that. It needs resilienacy.
ActivityPub uses URLs as IDs for everything. And there’s no way to update those IDs, it’s possible to update inbox URLs and other things but the main address of the object itself is its URL and thus there’s no way to propagate it without essentially making a new one.
It’s not impossible to do, but managing to get that to federate to all instances in a sane way is not currently possible.
There’s a ridiculous amount of URLs in the database and even fixing all of those won’t fully do the job, as post content might still refer to the old URL and whatnot.
It’s a messy situation, you’re not supposed to lose your domain.
p03locke@lemmy.dbzer0.com 1 year ago
Max_P@lemmy.max-p.me 1 year ago
You’ll have to go complain directly to the W3C for that. The situation is Lemmy may fix it with some custom protocol extensions, but then it’ll still break every other piece of software that follows the spec like Mastodon, Kbin and others.
It’s like adding a 6xx status to HTTP. You technically can, but expect every standard compliant clients to be confused and bail on it.
You can’t just change domains with emails either and have everything seemlessly migrate over. Not losing a domain is not a completely unreasonable assumption to make.
Thankfully the users and communities aren’t lost, it’s just that people outside of fmhy will have to resubscribe to the communities on the new domain.
Zetaphor@zemmy.cc 1 year ago
A lot of armchair developers in here who think there is an easy solution to distributed identity
Max_P@lemmy.max-p.me 1 year ago
There’s definitely better ways to handle this, like, the ID could be a public key or something. Chances of RSA/EC key conflicts is basically nonexistent or we wouldn’t use them.
But it’s the W3C, of course they assume URLs can and will be permanent. Your domain being seized is not something typical companies and organizations face. It’s something you expect to happen to a site hosting piracy and other illegal content, which FMHY is somewhat borderline with its piracy guides.
ActivityPub is not designed to be any sort of censorship resistant for sites that move addresses and servers frequently.
nintendiator@feddit.cl 1 year ago
Who was the idiot that decided to use for a database ID an identifier that almost entirely depends on external (and, for fediverse purposes, usually antagonistic) entities?
Max_P@lemmy.max-p.me 1 year ago
The W3C, apparently. It’s both the ID and the URL of the object if you want to refresh it. They seem to suggest doing it that way because the URL of a user profile is going to be guaranteed to be unique, and can only be owned by the owner of the domain.
Lemmy assigns it its own internal ID per instance but it’s only used internally for joins and stuff.
For example, your person ID is feddit.cl/u/nintendiator. If you curl it in ActivityPub format you’ll get your user:
nintendiator@feddit.cl 1 year ago
Immediate design issue right there: the URL of a user profile is not guaranteed to be unique, and while it can “”“only”“” be owned by the owner of the domain, 1.- it’s not owned by the user of the profile and 2.- the ownership by the domain owner is revocable by a third party.
Design-wise, it feels to me like they decided that land / house deeds could be certified by municipal traffic signage.
Zetaphor@zemmy.cc 1 year ago
The W3C, also known as the people who develop the web standards. It’s a reasonable expectation as you have to draw a line in the sand somewhere. Distributed identity is not a solved problem, so domains are the best solution we have right now.
What would you suggest they use as the identifier with which allows other entities uniquely identity you? There are no alternatives until you introduce a ton of cryptography, which is what DID hopes to address, but that’s still going to be bad UX.
nintendiator@feddit.cl 1 year ago
Figures. The same people who added DRM to the web standards.
Now, I don’t know what other alternatives could have been used, but I know that URL was among the obvious ones to not use. Something that uniquely identifies you has to be non-transitive and non-revocable by a third party, of which URLs are neither (domain names are revocable, URLs don’t have addressing persistence let alone when you add query strings into the mix, etc). Among the few things that I can think are non-transitive and non-thirdparty-revocable are the good ol’ ssh-keygen keys, easy to generate and all that but I’ve never found a good mechanism or design to query about them.