Quote:
Originally Posted by Kudbe
So ... relatedness is not inferred from the subject if other, rather technical cues are missing?
|
I don't know the exact logic, and I'm having trouble read the code to figure out (its deep in the Cyrus internals, and there's nobody in the office right now to explain it to me). But yes, subject handling has to be a bit magical when it comes to conversation construction, for two main reasons:
- Two matching subjects aren't necessarily part of the same conversation. Often automatic mailouts (notifications, bills) will have the same subject, even weeks apart. These are probably unrelated.
- Messages with matching References/In-Reply-To headers but different subjects often means that someone has replied to an old message and changed the subject, starting a new conversation. You don't want these to be grouped as they're unrelated.
Fortunately most sending clients do set References/In-Reply-To headers correctly, and most users do start a new message for a new topic rather than replying to an old one, so most of the time it comes out right. We do go over this in the office every couple of months, and so far what we have seems to be the most balanced approach.
I'll try to get the exact logic figured out sometime and post it. If Bron hasn't already posted it on this forum or somewhere else already, that is