565290 - Include greylist (or graylist) in spam filtering and handling

Status: RESOLVED WONTFIX

(Thunderbird :: General, enhancement)

Description

[Emerson Prado]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; pt-BR; rv:1.9.2.3) Gecko/20100401 Firefox/3.6.3
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; pt-BR; rv:1.9.1.9) Gecko/20100317 Lightning/1.0b1 Thunderbird/3.0.4

Greylisting seems to me a good way to minimize false negatives and positives. This way, only messages already known as good or spam will go to the specific folders. The unknown messages could be delayed as recommended in http://www.greylisting.org/ and, when actually received, put in a special folder ("suspect" or whatever). Then the user would set the unknown messages accordingly.
In short: an "unknown" folder and a delay (dunno if this delay is a mail manager's task).

Reproducible: Always

Comment 1

[Jo Hermans]

That's not how Bayes filtering works. Each message is assigned a score, but it can never be said if a message if 100% spam or 100% legitimate.

Comment 2

[Emerson Prado]

True. (I just googled Bayesian spam filtering to understand your reply, so read everything as from an absolute beginner)
But the software assumes every incoming message to be either spam or not. There is a threshold of spamicity, whatever one, which forks messages between black and white.
My suggestion is to have a grey range where most false positives or negatives are kept for the user to manually set its status. So it would afterwards be included in the training, as well as in the white or black list.

Comment 3

[Kent James (:rkent)]

You are confusing two issues.

"Greylisting" using the definition of your link is a function that is implemented by your email server. It cannot be done by a client. So if that is your request, then this bug is INVALID.

What you describe in comment 2 is some desire to manage junk emails by the degree of spamminess. That is not "greylisting". In Mozilla Mailnews, that degree of spamminess is called "junk percent". That is now stored in the core code, but there is very little user interface for it in the main application. Really the only thing that you can do with it there is to use it as a search term. So you could, for example, create a virtual folder that shows messages that are in a middle range of uncertainty.

My extension JunQuilla (https://addons.mozilla.org/en-US/thunderbird/addon/9886) exposes junk percent as a column variable, and adds an automatic Uncertain folder to support training on emails whose junk status is less than certain. Personally I find seeing the junk percent as enormously useful in quickly sorting spam from good emails. But people who have never used it don't see the value of it, so I've always encountered resistance at trying to expose it more in the core code.

Comment 4

[Emerson Prado]

> "Greylisting" using the definition of your link is a function that is
> implemented by your email server.

Thanks, I was really not certain of that. I suggested a similar thing in the mail client, but now I see "greylisting" can be a misnomer.

> My extension JunQuilla
> (https://addons.mozilla.org/en-US/thunderbird/addon/9886) exposes junk percent
> as a column variable, and adds an automatic Uncertain folder to support
> training on emails whose junk status is less than certain.

Ha! The "Uncertain" folder is exactly what I had in mind. I'll try it.

> But people who have never used it don't see the value of it, so I've
> always encountered resistance at trying to expose it more in the core code.

You're correct. Probably an average user would be confused. In my suggestion, I thought of that as an option, maybe even not a default one, directed to power users.
Since it's available as a plugin, it's already an option for power users. I'd change bug status right for you not even bother doing it, but I'm not sure if this is invalid, worksform or wontfix.
Thanks!

Comment 5

[Ludovic Hirlimann [:Usul]]

WONTFIX in COre for now as this is solved by a nice extension.