Hacker News Reader

Comment by mindslight | Hacker News Reader

It seems like you should be able to do a middle ground that preserves the commenter identity in each individual thread, while destroying the links across different threads.

For example, reserve the usernames 'deleted-xxxx' where x is a number. Then when someone wants their history obliterated, loop over every thread they have commented in. For each thread, choose a random number deleted-xxxx and assign that username to every comment of theirs in that thread.

Thus comments in the thread remain linked with one another, other deleted users in the thread remain distinct, and comments in one thread aren't linkable to a user's comments in another thread (perhaps lamentable, but required for unlinking spilled personal info in the general case)

(I chose random numbers rather than incrementing so that one can't start making inferences about users based on deletion order)

View on HN

It should just be randomly generated when the page loads. "Oh, this user has the 'dormant' flag set? Generate a uid on page load" so all their comments have consistency for that one viewer for that single page load. Then if you click that user to see all the posts by them, you get an empty set. (Have the search queries return nothing for dormant users.)

Both the null search and dynamic generation per pageload (or any other randomisation across accesses or sessions) would reveal which comments/posts have been anonymised. That might be a cue to an adversary to dig deeper.

That information might be possible to determine from a comprehensive archive of HN, but it would be much harder to obtain.

The ability to disown a specific set of content (I'm not sure HN permits this) would avoid that issue. The associated account would just have a limited history, not an empty one. That would be equivalent to an after-the-fact throwaway account, which is much less attention-grabbing.

HN will not do this.

The only thing they do is change your username on every comment you’ve ever made to something else. But they’re all the same , so you can still link all the comments to one identity, and if even one comment makes it clear it’s a real person, then they all are.

Requesting anything more results in them saying “oh yes, we respect users wishes, blah, blah” and no action.

> The ability to disown a specific set of content

We are willing to do this.

> Requesting anything more results in them saying “oh yes, we respect users wishes, blah, blah” and no action

We gladly work with users to redact any material that identifies a person or reveals their location, or that otherwise poses risk to their welfare.

As a third party observer, it feels like "work with" could still be hiding a lot of thorny differences in judgement over what constitutes material that should be removed. That's why I was outlining a broad-brush mechanical policy that would cover most of the cases (unlinking a user's contributions across different threads from each other).

The problem is that each case is different, and there's no one approach that would cover all of the cases. Indeed I can see that it doesn't actually address the concern in this case, now that I've been able to find out what the account is and investigate the issue.

More broadly, this is why it's extremely unfair to make public accusations the way this user has done in this thread. The way we handle these cases actually takes far more time and involves more effort and care than it would take to implement the approach being advocated here.

So then you should implement the approach being advocated here.

You said yourself it would take less effort, and the outcome is better.

Win, win.

Reply on news.ycombinator.com