Comment by xtrapolate
7 years ago
Do you think it's as easy as "rm"-ing a file away? Your data is kept internally in a multitude of different databases. Parts of it sitting in cold storage. Log files, caches. That data is split across thousands of different nodes. Each system has different data retention policies. Some databases don't permit removal of a specific record - the records must "expire" first. It really does take time to delete data.
True, but this isn't an excuse. It's slow to delete data because Facebook designed it that way. They could have designed for privacy and real-time deletion of data, but they didn't, because they didn't care.
> "They could have designed for privacy and real-time deletion of data"
Actually, they could not. If data is geo-replicated across multiple clusters, spread all over the place, divided into hot and cold storage layers - it's crystal clear you can't perform "real time deletion of data". Instantaneous deletion of all data, leaving no trace behind, can not happen under such complex constraints.
>> > "They could have designed for privacy and real-time deletion of data"
> Actually, they could not. If data is geo-replicated across multiple clusters, spread all over the place, divided into hot and cold storage layers - it's crystal clear you can't perform "real time deletion of data". Instantaneous deletion of all data, leaving no trace behind, can not happen under such complex constraints.
Yes, they could have. Your post is just a description of a design that can't delete data quickly. That doesn't prove that no design exists which can delete data quickly.
If Facebook had been designed with "we need to allow users to delete their data quickly and permanently" as a constraint from the beginning, it wouldn't look like the system you've described.
All you've done is pick all the things that Facebook did and say that if you do those things you can't delete data quickly. Yes, that's true--which is why Facebook would not have done those things if they cared about allowing users to delete their data.
1 reply →