Aahz Maruch aahz at
Thu Nov 25 00:43:33 EST 1999

In article <000801bf36c8$92ee19a0$292d153f at tim>,
Tim Peters <tim_one at> wrote:
>In practice, I like a two-step approach:
>1. Compute CRCs.  If they're different, the files are definitely different,
>   and you get out cheap.
>2. Compute SHA-1.  If they're different too, the files are also definitely
>   different.  If they're the same, then since the CRC and SHA-1 methods
>   are unrelated, it's reasonable to expect that the odds of a false match
>   are about 1 in 2**32 * 2**160 == 2**192 now.  This is even smaller than
>   the chance Python2 will have curly-brace block delimeters.

In practice, the likelihood of two files having the same MD5 or SHA is
nearly infinitesimal, provided they're the same length.  Check the first
hundred or thousand characters for equality and you're essentially
                      --- Aahz (

Androgynous poly kinky vanilla queer het    <*>
Hugs and backrubs -- I break Rule 6

Question authority: ask me anything!

More information about the Python-list mailing list