@badlogic woke up to a number of “yeah this is probably a caching policy at work” replies too. Seems a lot of people (including Mastodon) don’t know what a cache is if that’s the case.
Hint: if deleting the cache is a destructive operation that causes irrecoverable data loss, it’s not a cache, it’s your actual data.
@jwildeboer@badlogic the original post is in the context of a larger discussion that was/is about how, at least on our instance, any discussions that involves replies from people from other instances will eventually be broken because the posts from the “external” people will disappear, without those external people deleting their posts. This makes it impossible to refer back to older discussions.
You seem to be talking about active deletion of posts, where a user might decide to delete their own posts, or an instance might decide to delete its own posts. And I completely agree that, in that situation, those posts then being deleted across the Fediverse is a desirable feature.
@jwildeboer@badlogic But the problem (or at least what I perceive to be a problem) under discussion is not about active/intentional deletion of posts, but about something that presents itself as a caching policy, but breaks the decentralized nature of the protocol when enabled.
@jwildeboer@rovarma@badlogic in this case the original page is the one that's broken because of content retention settings. in this case the original, supposedly canonical view of the thread has 85% of the replies missing because the original server removed old replies because they were "old remote content", and in fact @rovarma is complaining that the only way to see most of those posts is to find someone who responded to the thread from a remote server, go to their server, and viewing the original post from there because that server never expired any of the content
this is clearly not a "working as intended" case and i'm baffled why you're trying to reframe this as somehow a good thing
@badlogic I remember Moxie saying online something along the lines of "No we are not going to do federation in Signal ever".
I just looked it up and did not know they were burned by the experience before.
Gultsch citing HTML/CSS/JS as a "federated success story" is extremely weird though. That's not a protocol and controlled by 3 big companies, as he himself says in his article.
@jabberati@neuron he's explicitely calling out HTML as the fast moving sucess story, while saying that HTTP is stuck. And HTML/CSS /JS is governed by a small cabal of (almost) trillion dollar corps.
@jabberati@neuron I don't see how web standards are themselves federation. They are client-side tech with HTTP as a transport layer. HTTP enables federation. But that is stuck in the 90ies.
@jabberati@badlogic@neuron Fediverse is stuck in time... It's just not old enough. Try changing the ActivityPub protocol - for example adding some cryptographic verifications to posts. You would have to convince all ~15 implementations to change which is practically impossible. You can update Mastodon, but Mastodon is not the whole Fediverse.
@jabberati@badlogic@neuron Yes, it is necessary in the long term, but totally useless if your app is laggy and no one uses it. Your resources are limited and you have to balance where to focus them. Signal decided to prioritize user-friendly app over decentralized design, which, in turn, has brought e2ee to many more people than XMPP or Matrix combined.
@badlogic Yeah sucks. But that is a sad point of how shit is. Unless you download all an keep it local stuff can always disappear. How many Twitter discussions I had with suddenly empty blocks because people quit or deleted their messages. Same shit different stink. Nothing is forever until you got a local copy
Add comment