Okay, what is the advantage of SHA1 over MD5, what does "crack" mean in that context?

Let me think "loud". The purpose of a hash is to group identical files or find more matching file for resume. Security is no issue here and wihle bad clients may theoretical lie, a later file test is done in the overlapping resume.
I think a hash should be follow some guidelines: a) be build fast b) should be unique enough to find identical files within a typical horizon (not high secure) c) should be identical even when people change metadata (eg MP3 id tag). So maybe it is enough to be build a hash over the first amount of MBs (skiping the very first bytes where some metatags are), even partials would be hit.

I don't know which algorithm will fit best to this guidelines. Anyone can help, has experience with different hashs?

CU, Moak