Most of the modern consensus #blocklists use tenforward.social as a trusted source and a generative source for blocking.
-
Many of the modern consensus #blocklists use tenforward.social as a source (direct or indirect) and a generative source.
I strongly advise against using any consensus blocklist that does so until the current situation with the admin is resolved. This is made worse because tenforward and .art often share block recs.
If you do use one as a source, IMO you should clearly communicate with your users what steps you are using to prevent false positives and citogenesis problems here.
-
If you run a consensus #blocklist I would suggest removing tenforward.social as a trusted source, recognizing that you are getting most of the same information from your other trusted sources (usually .art and r.l) as it is.
If you use them for _filtering_ (removal) purposes (you have a list and their list is simply used as a bias check) as opposed to additive purposes this of course doesn't apply.
If you manually review, please factor this in when you are doing that evaluation.
-
Oliphantom Menacereplied to Hrefna (DHC) last edited by
@hrefna Huh. It's not part of the TBS council, that I'm aware of.
https://tweaking.thebad.space/about
Gardenfence, yes.
Seirdy doesn't list the instances he uses for cross-referencing, I don't believe, but blocks have to exist on pleroma.envs.net first, so there's built-in protection there.
The Oliphant T0 list is now an amalgam of Seirdy's T0, Gardenfence, and IFTAS DNI list, so technically the Oliphant T0 list could include blocks sourced by tenforward.social.....
However, not just tenforward.social. Gardenfence requires nearly unanimous agreement.
"Suspensions which are shared by at least 6 of the 7 of reference servers are included in the list (~85%)"
-
Hrefna (DHC)replied to Hrefna (DHC) last edited by [email protected]
Note that this extends, though more weakly, to .art and r.l if they are sources for your blocklist: there's clear documentation right now of .art drawing their list directly from tenforward.social recommendations (via fediblock) without review or with only pro forma review, and r.l shares a moderator with .art (or did last I checked).
-
Hrefna (DHC)replied to Oliphantom Menace last edited by [email protected]
This concern extends to .art and r.l: https://hachyderm.io/@hrefna/113114357481398754
.art has been blocking, quickly, places recommended by tenforward.social without any apparent review (based on the notes on https://mastodon.art/about and what has been posted so far by the admin to tenforward.social).
So I'd argue that it is still an issue, albeit indirect for some lists.
-
@hrefna @oliphant Just for clarity as I'm not sure if I'm reading you right.
Are you saying that the issue is that using these instances is redundant (one copies another copies and another makes the consensus part not really consensus), or that there's something problematic with the decisions tenforward is making, or both?
I'm also missing what about the notes suggests no review. I tried spot checking but the defed notes looked normal and a lot of those recs were already blocked by .art.
-
Oliphantom Menacereplied to Hrefna (DHC) last edited by
@hrefna Perhaps, but Gardenfence is the only consensus list that has the sources you mention as overt contributors. Even as a block of three who might all agree on similar blocks for similar reasons (it's not unusual for councils to be talking about moderation issues offline), they still need 6 out of 7 to get any of those blocks on the list.
Individual servers like tenforward and .art are one thing, they can do what they like, but they aren't really voting to put anyone on a blocklist.
Instead, they are independently deciding, at the server level, to apply the block. That means they have to explain the block and negotiate with people on their own servers in the process, possibly pissing them off in the process. This is why .art has often announced a timeline on defederations, especially with larger servers, rather than throwing the switch immediately.
So basically, tenforward, .art, and r.l. did not, in fact, get together as a group and decide they are going to block and then went and did that.
Instead, a thing happen, they each apply a block, at at the end of the week, Gardenfence will tally up everyone's actual blocks, pulled from the server apis, and indirectly they will vote on the blocklist.
But 6 out of 7 will need to agree, and you'd need to watch the Gardenfence list itself, to really see if .art, tenforward, and r.l. are capable of pushing any new domains onto the blocklist, because if they did it means most of the rest of the consensus group agrees, too.
I'm not sure how much they are all talking or "colluding" offline, but it doesn't really matter. At the end of the day, if they apply a suspend on their server and cut off relationships, there will be consequences, people will move over it, etc. People may not even trust the Gardenfence blocklist anymore, depending on what kinds of stuff starts showing up on the list. But the consensus process works insofar as it reflects decisions already made, and high consensus means that unless you think the whole group is compromised, generally only what's truly urgent gets on the blocklist.
Now....also realize that mastodon.social is one of the servers on that list. So is indieweb.social (also federating with threads). They probably aren't going to block any of the servers involved, if I'm guessing correctly. Which means that none of the servers blocked by tenforward, r.l. or .art are going to end up on any blocklist at all, unless indieweb.social and mastodon.social decide to block them as well.
I mirror Gardenfence, so when it's updated you can see it in the commit history. So we'll see if anything like that comes to pass.
In my opinion, it's a very sensible list and I can't disagree with anything on it (except maybe a few 'dead' domains). That's one reason it's a reliable source as a standalone blocklist, or ingested as part of my T0 list.
-
Both. The admin to tenforward.social posted extremely violent fantasies directed at a specific individual and then proceeded to block anyone who called him on it.
Mastodon.art followed this by blocking many of the same instances. The note in mastodon.art for labyrinth.zone is racism and harassment, when this is the perspective from labyrinth.zone:
Vel :kyuuchan_deadinside: (@[email protected])
labyrinth zone meta, dot art, tenforward, general meta
(labyrinth.zone)
This is what was posted that triggered this result: https://tenforward.social/@packetcat/113099895540828598
-
Hrefna (DHC)replied to Oliphantom Menace last edited by
@oliphant Honestly I don't understand what point you are making here?
I do understand how consensus lists work and what they are or are not attesting to, as well as what individual servers can and cannot do. It doesn't remove the damage downstream from relying on lists that rely on flawed or compromised sources.
-
Oliphantom Menacereplied to Hrefna (DHC) last edited by
@hrefna I'm saying that there's only one consensus list I'm aware of that could be impacted (Gardenfence) and because it requires 6 out of 7 sources to agree, and since 2 of those sources are mastodon.social and indieweb, none of these blocks are getting on a consensus blocklist.
Unless you know of some other consensus blocklist with them on it.
-
Hrefna (DHC)replied to Oliphantom Menace last edited by [email protected]
@oliphant Indirect TBS, which draws from mastodon.art and r.l. Directly gardenfence, which as you point out draws from all three of m.a, r.l, and tf.s. Even with buffer, that should be considered compromised as a source due to the citogenesis problem.
-
Oliphantom Menacereplied to Hrefna (DHC) last edited by
@hrefna I don't include TBS in my T0, in part to avoid source pollution, the same sources from two different moderation councils contributing to the same amalgam list. With Seirdy (using basically his own blocklist, checked against consensus, which feels like a different thing) combined with Gardenfence (consensus) combined with the IFTAS DNI list (curated), there shouldn't be any overlap there.
But I think what you're getting as it that .art, tf and r.l. are all basically the "same source" and that's the citogenesis problem, even within a single blocklist project. I'm just.... not sure how being in agreement with each other makes all their past and future blocks suspect. Like they'll all come together to fight nazis and CSAM at the end of the day, despite whatever differences or biases, and that will be reflected on a high consensus list like Gardenfence.
But I'm not going to argue you into trusting the sources. If your ultimate issue is that the sources are compromised, so you can't trust anything that comes from them, I can't really argue with you. Quite the contrary. If you can't trust the maintainers producing the lists, you shouldn't use the lists, and that's always been the case.
-
@hrefna @ellesaurus @oliphant you are engaging in harrassment.
-
Then you are more than welcome to report me for it and I'll deal with it with moderation staff, should they judge it to be an issue. You are going to be blocked by me after this between this and your other post whining at me, so I'm not sure what you are hoping to gain here.
Considering your admin made a post about shooting people like me in the head with a russian firearm and that is "kind," I'd say my response has been rather tame, personally.