Hello All!

I know I need it spelt out nice and easy ;)

Someone (I'm not sure who, sorry), said that having a ^aHASH ,
of a message instead of a MSGID: was a bad idea.  

I don't understand why, I *think* the reasoning was that messages may get
grunged by intermediate sites, or that CRC isn't a reliable detection
method.

If a message gets grunged then it should be dropped as a bad message,
editing message text is baaaad.  If Origin's etc are modified then the
software is obviously broken that did it!

If the CRC isn't a reliable detection method, could someone please suggest
a more reliable one.  I don't study random-number theory, or cryptography
so I'm quite unaware as to what is a "good" and a "bad"
algorythm.  
CRC is a standard.  In my original proposal I suggested using two or more
CRC's and possibly a line count, thus making it as unique as possible.

Just in case here is my proposal:

* indicates a required step.  
+ indicates an optional step.
  indentation shows branches.

When you get a message:

 * Check for a ^aHASH kludge.
   * If it exists:
     + Check the HASH is correct.
        if not toss into Bad as a grunged message.
     * Check the HASH against "seen" hashes.  
         + if its the same do a full text compare if possible.
           + if the message isn't available export as normal, don't drop a
             message.
         * If it's the same toss as a dupe.
   * If it doesn't exist:
     * Dupe detect on MSGID
       * If there isn't a MSGID Dupe detect on header.
     * If it's not a dupe, add a ^aHASH and export as normal.

This will catch rescans, reuploaded qwk packets etc.  And If the HASH is
correctly choosen it can be depended upon to be 100% accurate if all the
optional steps are taken.  It would work even if many sites don't support
the ^aHASH kludge.

I'm sorry If I'm missing something blindingly obvious, or perhaps something
not so obvious to do with hashing theories.

---