@alice so well over twice as noisy on the network as deleting an account and all its posts with federated deletes, as it'd entail updating every participant server mentioned in every post or reply in the archive, yeah?

@djsundog i'm not sure the mentions would get much involved, unless you wanted to send a notifs again.
currently there's no way to update the post id at all, so the best you can do is make posts re-appear with a new id and never send any notifications about them unless you boost them. and remote instances would know about them as a new posts if they are interacted with. but then would keep both copies.
if we added a way to broadcast updates of all posts of an account to a new domain, it would be similar in noise as deleting an account

@alice I was thinking of the scale of notifying all servers with thread participants so they could update their references to the canonical locations of the newly moved posts, assuming their software cared enough to do so

@djsundog i mean it would be a less noisy to do this for the whole account and all associated posts in bulk, for everyone who saw them, like we do with account deletes, assuming we don't intend to support partial migrations.
going over every post and sending an update to its participants can be a lot yea, but that doesn't have to be done quickly (it's close the traffic automated post deletion would generate, which seems acceptable if done slowly)

@alice I agree, batching it in one bulk transaction would be the way to go for sure. the rest is just implementation details haha

@djsundog @alice so i have Thoughts here but none of them will ever be supported by mastodon et al and certainly not for old data

1) give each actor a namespace
2) let actors assign consistent identifiers within that namespace
3) let actors update their namespace when they change servers

all it takes is supporting relative references, which AS2 warns against as “too complicated :(“ so people have avoided it

@djsundog @alice honestly i am not even sure most fedi software supports redirects. i don’t see any reason why they shouldn’t, but i would not be surprised to find out that’s the case. who’s going to audit 100 codebases or spin up 100 test instances

@trwnh @alice I mean, given the option to support following redirects and rebuilding redirects poorly from first principles...

@djsundog @alice “rebuilding x poorly from first principles” fedi in a nutshell is2fg

@trwnh @alice the internet at large has a long and storied history of that so we inherited it naturally I suppose

@djsundog @alice “web architecture? what’s that?”

not that webarch is perfect or ideal but it is like The Thing That Mostly Works

linkrot is an issue for which our best solution is “just never change identifiers, get better foresight fivehead”