This is for loop checking, so the number of CRC signatures retained is the value of "Daily-Threshold=" (perhaps plus one). The CRC signatures are 32-bit numbers so the probability of false duplicate is very low. The header is not included in the CRC because, in a loop situation, the header is rarely identical on every iteration. CRC checking can be disabled with "Loopcheck= NoCRC".

  Eric

-----Original Message-----
From: LISTSERV List Owners' Forum [mailto:[log in to unmask]] On Behalf Of Hal Keen
Sent: Tuesday, March 07, 2017 16:16
To: [log in to unmask]
Subject: Loopcheck CRC and timing

Is there any detailed information about the CRC mechanism for loop checking? 
I'm interested in knowing the probability of an accidental match, and the retention time of CRCs being matched against new messages for duplicate detection.

This probably calls for some background: I manage lists for an IEEE standards working group. Drafts of standards under development are submitted periodically to the membership for voting and comment. We ask that responses be returned to a designated list, with a subject line identifying the draft and the vote. There's a ballot-response form, too, but if the vote is Approve without any comments, it doesn't collect any more information than is on the Subject and From lines of the mail header.

One subscriber decided that meant he could send his vote with no message body. That worked--the first time. The second time (a day later, on a different draft standard), the no-body matched the CRC of the first one, and loop detection rejected it.

This is not a huge problem; if he returns just the top lines of the form, it will have a unique document/draft designation which should make the message body unique. Easy fix (or would be, if he took advice more readily!)

But it does raise three questions:

1. Am I correct that the sender's address IS considered as part of duplicate detection?
2. What is the probability of an accidental CRC match between non-identical message bodies?
3. How long does LISTSERV retain CRCs of previous messages for duplicate detection?

Thanks for any enlightenment you can provide.

Hal Keen
 

############################

To unsubscribe from the LSTOWN-L list:
write to: mailto:[log in to unmask]
or click the following link:
http://peach.ease.lsoft.com/scripts/wa-PEACH.exe?SUBED1=LSTOWN-L&A=1

############################

To unsubscribe from the LSTOWN-L list:
write to: mailto:[log in to unmask]
or click the following link:
http://peach.ease.lsoft.com/scripts/wa-PEACH.exe?SUBED1=LSTOWN-L&A=1