Subj : Squish dupe detection.
To   : Marc Lewis
From : Mike Tripp
Date : Thu Nov 21 2002 09:56 am

Hello Marc.

20 Nov 02 20:29, Marc Lewis wrote to Mike Tripp:

 ML> I also have a problem with that.  I post messages to the Vatican and
 ML> the Intelec IN_Catholic echos with a batch file (using MsgPost/TE)
 ML> that waits 8 seconds between each post, guaranteeing a distinct MSGID
 ML> for every message.

If you enable the header hash instead of MSGID, Squish's granularity is 2 secs 
on the timestamp.

 ML> At this end (at my point) I have ONLY MSGID enabled in the
 ML> squish.cfg.
 ML> It tosses the Vatican messages and traps the identical messages (with
 ML> different MSGID's) that should post to the IN_Catholic echo. (Which is
 ML> in a different Zone as well.)  _That_ is broken.

Yep...it also sounds like an old 1.10 bug that was supposedly cleaned up in 
1.11.  Might need to seek and destroy an old copy of MSGAPI(32).DLL?

=== Cut ===


              SQUISH 1.11 REVISION NOTES -- SUMMARY OF CHANGES FROM 1.10

[...]

          *    Added  the "DupeLongHeader"  keyword.   If  this keyword  is
               enabled,  Squish  will  use  the entire  subject  line  when
               performing  header-based duplicate  checking.   Without this
               keyword,  Squish will check only  the first 24 characters of
               the subject field when performing duplicate checking.
[...]


In addition, the following problems from version 1.10 were corrected:

[...]
          x    Squish  will no  longer crash  if "Duplicates  0" is  set in
               SQUISH.CFG.

          x    Squish will  now properly  process the  "DupeCheck" keyword.
               Previously,  Squish was  always using  the MSGID  and header
               algorithms, regardless of the setting in SQUISH.CFG.
[...]

          x    Fixed  bug that  caused  Squish to  incorrectly parse  MSGID
               lines that contained double quotes.

=== Cut ===

 ML> In addition to the different MSGID's, the messages have the same text
 ML> in the Subj: area, but spaced and punctuated differently in 4
 ML> variations, just to break things up.  That makes no difference
 ML> either.

Which, of course, will only help if you're dupechecking header hashes and if 
the variance is more than 24 chars out, you'd additionally need 
"DupeLongHeader".

 MT>> It is the messages that are broken...not the tosser.  If you want
 MT>> a hack, hack the messages into compliance, so that Squish will
 MT>> pass them.

 ML> I will differ with you on this in this particular scenario, Mike.  I
 ML> sincerely believe there *is* a logic defect in Squish's dupe detection
 ML> routines.   Hopefully someone will find it and correct it some day.

Agreed, if you've got MSGID-checking only and all MSGIDs are unique.

Mike

--- GoldED 2.50+
 * Origin:  TechKnowledgy at Work              (1:382/61.1)

.