Body Checksum Calculation

Vernon Schryver
Mon Apr 1 18:34:54 UTC 2002

> From: "Tony L. Svanstrom" <>

>  Without having had the time to check the algorithms used (they don't seem to
> be easily accessable on the website)

Please do not discuss the fuzziness of the checksums in public,
to postpone the day when the next revisions will be required,

>                                      I would say that the problem is MUAs that
> change Content-Transfer-Encoding: and similar things; maybe even mess around
> with multipart-messages.

I have reasons for hoping that is not a problem with the DCC.

>  Personally I'm right now using SpamAssassin and DCC together; where SA is
> "best" one, and DCC is catching a few that SA don't. In the future I intend to
> check e-mails with DCC first, and then report e-mails to DCC (whatever system
> that wants to listen, and a private one) only if they are caught by SA,
> DCC-value of less than [some number] and maybe some close to foolproof tells
> that apply to my situation only.

That may conflict with the fundamental idea of the DCC.  That idea
is that by reporting cryptographic checksums of lots of mail to
clearinghouses regardless of whether it matches some pattern or
came from some blacklisted source, bulk mail can be detected by
all except the first first few targets.

Spam traps are useful only for defending those first few targets.  As
far as knowing whether a message is "bulk," it does not matter whether
the DCC clearinghouse says its body checksums have counts of 100 or
16,777,216.  You know that at least one other target received the
message that the sum of the reported target counts mark the message
as "bulk."  If the source or other characteristcs are not in your
white list, then it must be "unsolicited bulk" or "spam."

(The DCC count of "Many" means no more and no less than
"this checksum's count of targets is at least 16,777,216.)

Vernon Schryver

More information about the DCC mailing list

Contact by mail or use the form.