Yellow Swordfish said
The process of collecting emails and turning them into forum posts is easy. That's done and being tested now.

The difficult part is working out the best and most foolproof way to identify the sender of the email to ensure they are allowed to do this and to identify the topic and forum they want to post in. And this is made far more complex if they want to be able to start new topics.

When we resolve those issues and have a secure and workable model then the coding is the easy part.

Regarding the 'to identify the topic and forum they want to post in', for the sake of discussion, let's consider the case of replying to a new post in a subscribed forum or topic. I think, for my purposes anyway, that it is reasonable to assert that for someone to reply via email, they must be subscribed via email. In other words, they must receive an email from the forum that they may reply to.

The email they receive (on the subscribed topic or forum) would contain, if I were coding it, several sets of (unobtrusive) markers to identify the topic and the intended recipient. That way when the user replies to that email, we hope that at least one set of identifying marks makes it through unscathed (checksummed, of course) and that the surviving markers agree with one another, etc.

As to identifying them, that is trickier since mail From: addresses are so easily spoofed of course. That is why the markers contain information identifying both the recipient and the message. That way if an attacker pretends to be someone else and uses a forum member's email address to attempt to post, it won't work because there would be no markers. Or if there were markers, they wouldn't match the markers the legitimate user would've had in his message.

In other words, a single Post P where User A and User B both subscribed to the forum. User A's email would contain the contents of the Post plus multiple marker sets, all of which identify it as an email about post #P sent to UserA.

User B would receive the same email, except that his markers would be different: Post P + UserB.

So if User B replies to User B's message From: User B's email account, the software knows it came from User B as a legitimate reply from User B. An attacker User X could not post to the forum via User B's email account without also having the actual email that was sent to User B--which would mean that User X has already hacked User B's email account, and is not our problem. 🙂

Sorry if this is too much...just got to thinking through the security aspect and thought I would offer it up.

The markers could be as simple as PostID XOR UserID and then run it through a one-way hash. Because this would generate unique markers for every post, this would be a lot of trouble for someone to game for spam purposes.

Note this doesn't solve how to allow for brand new Topics via email at all, which is a tougher problem, I agree. For my purposes, for some time at least, that is okay. Replying via email is 80% of the value of this feature anyway. New Topics can wait for later.

Thanks!