serv/mox
1
0
Fork 0
mirror of https://github.com/mjl-/mox.git synced 2025-01-28 23:35:56 +03:00
Code Issues Projects Releases Packages Wiki Activity Actions

improve comments

Browse source
This commit is contained in:
Mechiel Lukkien 2023-07-23 09:42:29 +02:00
parent 70806137da
commit 4a4d337ab4
No known key found for this signature in database
2 changed files with 33 additions and 31 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

20
smtpserver/reputation.go
View file

@ -35,22 +35,23 @@ const (
//
// The decision is made based on historic messages delivered to the same
// destination mailbox, MailboxOrigID. Because each mailbox may have a different
// accept policy, for example mailing lists with an SPF mailfrom allow. We only use
// messages that have been marked as read. We expect users to mark junk messages as
// such when they read it. And to keep it in their inbox, regular trash or archive
// if it is not.
// accept policy. We only use messages that have been marked as either junk or
// non-junk. We help users by automatically marking them as non-junk when moving to
// certain folders in the default config (e.g. the archive folder). We expect users
// to mark junk messages as such when they read it. And to keep it in their inbox,
// regular trash or archive if it is not.
//
// The basic idea is to keep accepting messages that were accepted in the past, and
// keep rejecting those that were rejected. This is relatively easy to check if
// mail passes SPF and/or DKIM with Message-From alignment. Regular email from
// known people will be let in. But spammers are trickier. They will use new
// known people will be let in. But spammers are trickier. They will use new IPs,
// (sub)domains, no or newly created SPF and/or DKIM identifiers, new localparts,
// etc. This function likely ends up returning "inconclusive" for such emails. The
// junkfilter will have to take care of a final decision.
//
// In case of doubt, it doesn't hurt much to accept another mail that a user has
// communicated successfully with in the past. If the most recent message is marked
// as junk that could have happened accidental. If another message is let in, and
// as junk that could have happened accidentally. If another message is let in, and
// it is again junk, future messages will be rejected.
//
// Actual spammers will probably try to use identifiers, i.e. (sub)domain, dkim/spf
@ -60,7 +61,7 @@ const (
// Some profiles of first-time senders:
//
// - Individuals. They can typically get past the junkfilter if needed.
// - Transaction emails. They should get past the junkfilter. If they use one of
// - Transactional emails. They should get past the junkfilter. If they use one of
// the larger email service providers, their reputation could help. If the
// junkfilter rejects the message, users can recover the message from the Rejects
// mailbox. The first message is typically initiated by a user, e.g. by registering.
@ -69,7 +70,8 @@ const (
// domain that would let the message through.
// - Mailing list. May get past the junkfilter. If delivery is to a separate
// mailbox, the junkfilter will let it in because of little history. Long enough to
// build reputation based on DKIM/SPF signals.
// build reputation based on DKIM/SPF signals. Users are best off to
// configure accept rules for messages from mailing lists.
//
// The decision-making process looks at historic messages. The following properties
// are checked until matching messages are found. If they are found, a decision is
@ -86,7 +88,7 @@ const (
// domain instead of the exact domain.
// - Matching DKIM domains and a matching SPF mailfrom, or mailfrom domain, or ehlo
// domain.
// - "Exact" IP, or nearby IPs (/24 or /48).
// - "Exact" IP, or nearby IPs.
//
// References:
// ../rfc/5863

6
store/account.go
View file

@ -364,9 +364,9 @@ func (m *Message) JunkFlagsForMailbox(mailbox string, conf config.Account) {
// first-time incoming replies from users this account has sent messages to. On
// IMAP append to Sent, the message is parsed and recipients are inserted as
// recipient. Recipients are never removed other than for removing the message. On
// IMAP move/copy, recipients aren't modified either. don't modify anything either.
// This works by the assumption that an IMAP client simply appends messages to the
// Sent mailbox (as opposed to copying messages from some place).
// IMAP move/copy, recipients aren't modified either. This assumes an IMAP client
// simply appends messages to the Sent mailbox (as opposed to copying messages from
// some place).
type Recipient struct {
ID int64
MessageID int64 `bstore:"nonzero,ref Message"` // Ref gives it its own index, useful for fast removal as well.