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
While there have been well publicized problems with MD5 due to collisions, UNINTENTIONAL collisions among random data are exceedingly rare. On the other hand, if you are hashing on the file name, that's not random data, and I would expect collisions quickly.
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.