md5 and large files

["Martin v. Löwis"]

Brad Tilley wrote:
> All you need is 2^128+1 to find a duplicate, no? The problem, as I 
> understand it, is getting to the end (2^128+1) as sufficient computing 
> power isn't available... yet.

No. After 2^128+1, you will have found two files with the same md5sum.
However, you won't necessarily have found a file that hashes the same
as your *original* file. You may well search the space of all possible
files (in all file lengths), and never find a single file that hashes
the same as your original file.

Regards,
Martin