Facebook doesn't delete photos when removed, seems to keep them indefinitely | Hacker News Reader

Facebook doesn't delete photos when removed, seems to keep them indefinitely

photos-a.ak.fbcdn.net

15 pointsandreyf10 comments

I'll look into what's happening, but in general it's expensive / complicated to delete from edge-caching CDNs.

If there is a photo that gets constant traffic, the edge caches will do their job of keeping it around. Poking around the Googles, it looks like this URL has been reposted a lot, which creates a self-fulfilling prophecy and will keep it around in cache.

http://www.google.com/search?q=5013_96426961157_709851157_21...

So you can't delete delete a picture if other people, who don't own it, like to look at it?

Not very hard to imagine how that might be exploited in some undesirable ways.

It's more complicated than that and I'm talking and about edge caches in general here, not for my employer. I once saw a Flickr employee (Cal Henderson?) pluck the original Ludicorp logo out of Yahoo's edge caches. That's not to mention ISP caches and proxies, third party image search, etc, that neither you, I, Akamai, nor Facebook control. I'll look into it and see what's going on in this case.

I recall reading that the way the images are stored (for efficiency's sake) makes their deletion difficult. This analogy may not be all that correct, but I think of them as part of an archive in a fashion where their removal means rewriting part if not all of the much larger archive.

This is probably too simplistic, but the information caused me to wonder whether it would be practical to overwrite them in-place with equivalently sized data that represent a pure white or black image containing no metadata (EXIF, etc.). Data traces might be left, e.g. file name, perhaps some indexing (are metadata indexed?), but the image itself would essentially be gone.

http://sphotos.ak.fbcdn.net/hphotos-ak-ash1/hs343.ash1/29302... (this picture is to make a point, but it was just uploaded to FB then deleted, I think this guy has a point)

Hm, wasn't Facebook launched in 2004?

Exactly. I can put whatever date I want onto an image, thus we should view the duration here with skepticism, but this does appear to be an issue nonetheless.

Facebook has talked about this before in their tech blogs or presentations, mentioning it was too expensive to actually run deletes the way they store things; and that it was actually "cheaper" and easier to just keep going and not look back.

It's probably because of how they store image data. http://www.facebook.com/note.php?note_id=76191543919

Reply on news.ycombinator.com