3 ms·
Regardless of beliefs on the Right to be Forgotten, this feels wrong. The way that I had previously interpreted it, a person or entity could ask to have all re
by robinson7d 11y ago
Regardless of beliefs on the Right to be Forgotten, this feels wrong.
The way that I had previously interpreted it, a person or entity could ask to have all references to them removed. The request in this article isn't that, though - it's a request to cherry-pick, and have subsets removed.
To me, this feels too exploitable. I'm not sure whether I could agree with a Right to be Forgotten in the first place, but if there were one it should be all-or-nothing.
---
EDIT: It appears, per @alain94040's comment, that this is not entirely the case, and that the article was not accurate.
The interpretation that @alain94040 explained, which is that Google would still show the articles in question, just not if the query includes the entity who requested they be forgotten's name (or is it just not weighted by the name? If a name's included, but the rest of the query is sufficiently strong, would it still find the article?) actually seems completely acceptable in the realm of a Right to be Forgotten.
- ergothus 11y agoI think this demonstrates the problems with restricting information (Note: Not making a moral judgement of any kind, just stating the problem set is HARD) Information has basically zero copy cost, and can't be removed once given. So to stop people from knowing, you have to stop them from originally learning. Except that the ability to learn is "cheap". The same problems result whether we're preventing the spread of how to build a bomb, breed a superbug, or remove naked photos. Somewhere in that continuum will be info that you agree is best not to be spread, but the steps required to enforce that can be a lot more expansive. It's hard.