eval legit or junk?

POP Peeper: Tech support, suggestions, discussion, etc.
Post Reply
gordon
Posts: 208
Joined: Tue Sep 02, 2008 6:43 pm

eval legit or junk?

Post by gordon »

Hope you can see this.
This is interesting.
Check mail is fresh.
This message is colored red, but the evaluation says legit (-2.5)
Also, doesn't a thumb down flag cause an "Eval Junk", overriding everything?
This is a .ru junk mail.

There were 2 recipients on this mail.
Should multiple recipients activate a rule?
Gordon
Attachments
PP1.jpg
User avatar
Jeff
Admin / Developer
Posts: 9227
Joined: Sat Sep 08, 2001 9:46 pm

Re: eval legit or junk?

Post by Jeff »

As far as the coloring is concerned: the thumbs down (meaning, you've deliberately specified it as junk) overrides the evaluation. After you mark the message as Junk, you could re-evaluate the message and it's very likely that it will be evaluated as Junk; but PP does not re-evaluate the message automatically after you mark it as Junk/Legit, so you're still seeing the original evaluation in your example. You can re-evaluate it by right-clicking and selecting AntiJunk -> Re-evaluate message (or under the main 'Messages/AntiJunk' menu, or under 'AntiJunk' when viewing the message).


Multiple recipients does not have any effect by default; multiple recipients is not uncommon for business, friends, family, etc.

If you want to add a check for multiple recipients, this is how it could be done (but note that this isn't perfect):

Code: Select all

[Rules]
Multiple recipients_Action = 0 0 0 "" "" 0 16777215 ""
Multiple recipients_Rule00 = 0 0 3 "To" 0 ".*@.*@"
Multiple recipients_Status = 
Copy that and then create a new rule in POP Peeper and select from the menu: Edit / Paste Rule Set. Note that this will have the action of evaluating the message as "Junk"; you can change that if that's what you want.

A couple things about this rule:

1) What this rule does is check if there are at least 2 '@' characters in the "To" field. This is not perfect because it can trigger messages that only have 1 email address, but with a name containing an @. ie. it's not entirely unheard of for companies that don't know your name to replace the name part of the email field with your email address, e.g. "test@test.com" <test@test.com>. There may be fancier ways to mitigate this, but regex is not my forte and I'm really only posting this as a "if you really want to..." and not as a recommendation that it should be used.

2) This rule specifically only checks the 'to' field and does not use the "+To/CC" option that's available. The reason for this is that while I was testing the above rule, I discovered that there's a bug that prevents regex (and wildcard) from working on the "+To/CC" field (and only this field). This will be fixed in v5.1.4.
gordon
Posts: 208
Joined: Tue Sep 02, 2008 6:43 pm

Re: eval legit or junk?

Post by gordon »

Thanks for the explanation. I did not know that manual evaluating was necessary.
I assumed any necessary evaluating would be updated at "check mail".

1) I wish that PP automatically evaluates after a thumbs down/up change.

2) You can see the information popup window has no "User: junk/legit"
Shouldn't that have "User: junk/legit"

3) In the attached the evaluation column rectangle is green, but the Evaluation box
and message are red.
They should all be the same color, right?
Attachments
PP2.jpg
User avatar
Jeff
Admin / Developer
Posts: 9227
Joined: Sat Sep 08, 2001 9:46 pm

Re: eval legit or junk?

Post by Jeff »

gordon wrote: Fri Sep 17, 2021 2:24 pm 1) I wish that PP automatically evaluates after a thumbs down/up change.
My initial response to this was simply going to be that, after a while when you've marked enough messages, it doesn't really matter. And my main hesitation is that evaluating a message isn't resource-free.

BUT... most of those "resources" are in the URI/DNS-BL's, so if PP were to just re-process the affected filters (bayesian and white/black list) and simply re-use the previous results for the others, then that wouldn't be too bad. So... yeah, added to my list of things to consider.
gordon wrote: Fri Sep 17, 2021 2:24 pm 2) You can see the information popup window has no "User: junk/legit"
Shouldn't that have "User: junk/legit"
It could (and I added that to my list, too), but I don't think it's particularly relevant -- The purpose of that popup is to show why POP Peeper evaluated the message the way that it did, e.g. if a rule mis-evaluated a message, then you may want to tweak or disable that rule.

gordon wrote: Fri Sep 17, 2021 2:24 pm 3) In the attached the evaluation column rectangle diamond is green, but the Evaluation box
and message are red.
They should all be the same color, right?
No. The diamond represents how POP Peeper (aka "system") evaluated it -- it thought it was legit. The thumbs down represents how you evaluated it -- you said it was junk. As mentioned before, as far as coloring is concerned (and, by that, I mean the background coloring), the user's evaluation trumps the system's evaluation, so that's why it's colored red. The contrasting diamond color could be used to quickly see which messages are disputed and allow the user to investigate further if they want.


In this specific case, even re-evaluating the message won't matter because the (edit)Bayesian filter has already determined it was Junk. But, the email address is in your white list, which has more weight than Bayesian. I would say that it's a questionable tactic to mix how you mark messages from one sender. If you're signed up for a newsletter and there are particular types of email from that sender that you're not interested in, it's usually better to create a specific rule to delete those (or see if you can opt out of those specific types of email). There are probably other tactics you could try if it's for curiosity sake (change the weight of the white list to 1.0, or remove their address from your whitelist -- actually, maybe not removing because it could be re-added to a white/black list when you mark it, so maybe change the weight for that specific email address to 0.0... ah, you can't do that because it thinks you want to use default in that case, so maybe 0.1 [lower values are quirky and may or may not work, added to my notes]).
Post Reply