Comment 132 for bug 1546507

So the problem with trying to solve this by just preventing creation is that it fixes the issue only partially. All the images that might have this situation already in hands (People creating image based on public image's location to have their own metadata associated with it) are still vulnerable as soon as any of these images with multiple same location is deleted. In other words the creation time fix is not a solution for any existing cases where the reference counting would be.

Now if I have understood correctly one cannot remove snapshotted image from rdb backends, which makes this bug not disappear but be bit less hazardous. I will try to test that next week.