@meso@fba There is definitely no CP at that domain. There is nothing at that domain but a notice that the instance is closed forever. I would also not trust Ze Jahmans to decide what does or does not constitute CP.
Hey, @mint, how difficult would it be to do a "top schizos" list? Like, the instances that block the largest number of servers that were dead before the instance existed? I will add endpoints to fedilist to support this effort.
@p@NonPlayableClown@fba >Like, the instances that block the largest number of servers that were dead before the instance existed? Might be possible, but would require some complicated queries and I know jack shit about SQL. @pwm made https://fba.ryona.agency/weighted which makes solves the problem of the opposite end.
Speaking of schizos, there's a guy that tired maknig a big deal of reordering the domain of an instance that barely amassed 100 posts in all three months it existed. Screenshot_20240722_204201.png
@p@fba They've supposedly closed down on 1st of July, but they're still spamming Delete requests over the whole network, making my instance fetch their empty profiles (at least until I patched pleromer to MRF check deletes earlier in pipeline). Noticed that the funny block reason didn't get fetched by FBA since the instance was probably offline when fetching peerlists, so decided to grab it manually while it's still up.
@p@NonPlayableClown@fba@mint I don't think it's that schizo. I copied whatever kf.cc defederated. I know they really were nasty pedos. I'm not gonna check if they are still there.
@Hoss@p@fba No, it's just that fetching (empty) actor that deleted some post comes earlier in pipeline than rejecting the delete activity. This adds some bloat to users table.
@goatmeal@p@NonPlayableClown@fba Reasonable, but I think what he's talking about is grabbing as many blocklists as possible, no matter how obsolete or incorrect those might be.
@mint@p@NonPlayableClown@fba I can definitely cook up a beating a dead horse winrar list if there's some interest. First impression tells me that we would need an api to fedilist that asks if an instance is dead, and if so, since when. Then every time you record a block you make sure you know whether the blocked instance is alive or dead at the time of the block according to fedilist, and update your database accordingly if you have no data. Probably need some contingencies for resurrection (fba-side), and a minimum time after which someone is dead (fedilist side), and some caching for live instances to make sure fedilist isn't hammered with queries about if poast is dead yet (fba-side). After that the query is simple. It's intelligently ascertaining the information and keeping it in sync that's a bit weird but not impossible.
> api to fedilist that asks if an instance is dead, and if so, since when.
Yeah, the status changes page would normally do that, but what is actually useful in this case would be the last successful contact timestamp. That's on the page, I can make it an endpoint, tack on something like the time it was first discovered, the most recent successful contact, etc. It'd probably be better to do this by putting microdata annotations on the page (so you just request the regular page and then pull out the timestamp), but for the lulz I am willing to do a JSON endpoint.
Speaking of status changes, I've been meaning to make the status changes page accept multiple hostnames because that makes the RSS feed a little more useful, you can subscribe to a set instead of subscribing to 50 RSS feeds for individual instances, so that's done, it's one of those things that is trivial enough that it always slips my mind: http://demo.fedilist.com/instance/recent-changes?host=fsebugoutzone.org,freespeechextremist.com,shitposter.world,gh0st.live,ryona.agency,poa.st . I think kirby was interested in that but I can't remember on any given day what server kirby is on.
> Then every time you record a block you make sure you know whether the blocked instance is alive or dead at the time of the block
Yeah, that was my thinking, but that it'd be recorded locally in the DB, so that someone imports a blocklist and then you don't need to fetch a thousand instances every time. Then when you do the query, you just check, like `where block_timestamp > last_alive_timestamp` or if you want to be certain, `where block_timestamp > last_alive_timestamp + '30 days'::interval`. Something like that.
> a minimum time after which someone is dead (fedilist side),
It's marked as "down" if it's been at least an hour and there have been at least four unsuccessful attempts, but I think if I just write the timestamp of the last successful contact, that's a little more flexible. Like, you can get the status by just hitting the recent changes CSV: http://demo.fedilist.com/instance/recent-changes/csv?host=fsebugoutzone.org.
> Yeah, that was my thinking, but that it'd be recorded locally in the DB, so that someone imports a blocklist and then you don't need to fetch a thousand instances every time.
Sorry, poor phrasing, that was also my thinking. You could just keep a table of "instance, state, last_checked" and refresh that record if last checked was past some arbitrary threshold. Hell you could even do exponential backoff on checks with a counter column (reset the counter to 0 when the state is toggled). Would still allow for revival of dead instances (does that ever even happen though?).