No, he calls it slowbanning. Hellbanning is when you let them keep posting but other people can’t see the posts. Hackernews hellbans as well and you can see the comments from the deplorables if you turn on “showdead” in your profile.
IIRC hellbanning is a variant of shadowbanning that prevents easy discovery of the banned status by putting all suspect accounts into the same "hell" invisible to normal users.
This works well if you have a network of fake accounts from a single "persona" or ring of personas - by all their indications they can't see their own posts are being ignored.
Notice: it's almost the exact response to the persona management software problem [1] (aka bots).
This also works incredibly well for cheaters in videogames.
Give them their own queue with their own games with other cheaters to play against, and as long as nobody is cheating in a way that breaks the servers, they can play their own version of the game if they want without ruining the game for those who don't cheat.
This reminds me of someone telling me that they are using cheat sites for an online Scrabble game because they suspected their opponent (a “friend”) was cheating. It’s hilarious to think that two humans are watching two instances of a likely optimal bot play against each other and rooting for their instance of the bot.
It doesn't work for logged-out users. If I can just look at e.g. the Internet Archive's copy of reddit and see if my accounts are in there, it defeats the purpose.
Neither reddit nor HN make any attempts to make it hard for sophisticated users to figure out they're shadowbanned.
Forcing you to query IA at least reduces the frequency of feedback, as they're taking snapshots instead of giving a live feed. You could also shadowban IA. You can also do things like guess based on IP address or browser fingerprinting, or require a login from various IP ranges.
Of course your main point - that this is all terribly imperfect and won't stop a determined, sophisticated user, who has realized what's happening - is spot on. That, however, is perhaps a rare combination, rare enough to simply continue dealing with manually.
The feedback doesn't have to be very fast. If I'm botting correctly, my accounts will almost never be banned. Even once every 24 hours will be more than enough.
IA was just an example, and Tor would be easier. But anyway, I think it shows the flaw in doing so:
> You could also shadowban IA.
If the spammer manages to get all the IPs hellbanned just by looking at things, he gets more eyeballs on his spam.
My point is, you can't get much better than normal shadowbans, which are trivially detectable for moderately sophisticated users (just log out and try to check your profile) but not anyone else. "Hellbanning" is a stupid extension of this concept which only works in video games.
Also, shadowbanning is a spineless and deeply unethical move. If I get banned, I know what I did wrong and can reflect on that. If I get shadowbanned, I'm just screaming into the ether. That is not a Good Thing™, it is atrocious.
We only do that when an account doesn't have much history on HN and there's evidence of spamming or trolling. For established accounts, we tell people we've banned them and why.
I've always thought people might catch on, since no one is engaging with their takes. I'm curious if letting bots do some markov responses might keep them in the dark a little longer.
I see it far less frequently now (thanks @dang!), but a few years ago it wasn't uncommon to see shadow banned HN users continuing to post for years, talking into the void. Sometimes I'd look into their post history and so many of them had been banned for utterly trivial reasons, it was pretty sad.
Would you have a profile or two handy that I can take a look at? I left showdead on for a while but found it useless in terms of coming across interesting comments from such users. Thx.
I seem to remember being pretty upset when I found out Terry Davis of Temple OS was being shadowbanned. Maybe it was just temporary for a short period or a single comment, but I didn't like it.
Terry was permanently shadow-banned, the subject came up quite often. He suffered from schizophrenia and would make racist and paranoid comments about 80% of the time, but his remaining posts were often pretty technical / insightful. As the victim of an illness it seemed unkind that he was banned, but at the same time a lot of his output was obviously very offensive.
If I see an account that's been shadowbanned for years, and has consistently posted appropriately for maybe several months, I report it via the contact address.
I think nowadays you might catch on because the culpit might not be identified to your site with both their phone and their PC or something among these lines.
That's not true. Software filters sometimes kill such comments, but the accounts themselves are unaffected, and moderators review the killed comments and unkill the clearly good ones, and mark such accounts legit to immunize their future posts from those filters. Also, users often vouch for comments that have been killed in this way, which restores them.
My main issue was with the deplorables comment. I actually quite like the HN system of vouching.
> and mark such accounts legit to immunize their future posts from those filters
I remember seeing you restore a post from someone who made their account via tor and their comments were auto-deleted. Their next post was autodeleted in the same way, so I presume that this feature is buggy (or was, as this was quite a while ago).
There might be other explanations. For example, if an account shows signs of being connected to previous banned accounts, we might unkill a good post, but not immunize the account overall, until it establishes more of a track record.
I'm glad you like the vouching system! I still feel like it's the best single change we've made to HN since pg retired.
“Tarpitting” is a specific type of network-layer defence which is not related to degradation of service. A tarpit will typically stretch out response time to network-layer and some application-layer communications in order to waste wallclock time of spammers.