Re: [Pan-users] References field broken

[Charles Kerr]

On Wed, Apr 10, 2002 at 04:24:59AM +0200, Colin Leroy wrote:
> On Tue, 09 Apr 2002 18:22:03 -0600 Jim Henderson <address@hidden>
> wrote:
> 
> > There was a discussion about this last January, and it was decided that
> > pan shouldn't validate the references header because it breaks
> > threading.
> 
> It seems to be ok :)
> 
> > Charles, your message to me and the list read:
> >
> > "I'm committing a patch to fix this.  I'm changing the
> > broken-message-id-removal code to not remove the message-ids that we
> > need for threading (the first id and the last three) from the
> > References: header, even if they are syntactically invalid."
> 
> Well, the removed ones here were the last ones :-/

Yes, Pan regressed on this at some point.  I've refixed this and
added a regression test into pan/tests/test-gnksa to make sure it
stays fixed this time.

The comment for the new version of gnksa_check_message_id reads:

/**
 * Validates a Message-ID from a References: header.
 *
 * The purpose of this function is to remove severely broken,
 * usually suntactically-invalid Message-IDs, such as those
 * missing '<', '@', or '>'. 
 *
 * However we want to retain Message-IDs that might be sloppy,
 * such as ones that have possibly-invalid domains.
 *
 * This balance is from wanting to adhere to GNKSA that wants us
 * to remove `damaged' Message-IDs, but we want to be pretty
 * forgiving because these References are required for threading
 * to work properly.
 */