I\'ve got an image library on Amazon S3. For each image, I md5 the source URL on my server plus a timestamp to get a unique filename. Since S3 can\'t have subdirectories, I
Doesn't really matter how likely it is; it is possible. It could happen on the first two things you hash (very unlikely, but possible), so you'll need to support collisions from the beginning.