Dealing with Email Bounces in PHP
I do work for some websites that send email to customers. Order receipts, customer support, periodic newsletters, etc. I became an email administrator without knowing it. Now I have to find ways to automatically deal with email that bounces. That can happen for various reasons, like a user having an invalid address, being over quota, or blocking a message with a whitelist/overly-agressive spam filter (my clients aren’t spammers).
I have things set up to bounce most errors to a special bounce email account. You can do this by adding an Errors-To header in your messages. This way I have them all in one nice email account, and I can try to process the errors automatically. I set that up a couple years ago, and I’ve since found out that there are tons of different ways for mail servers to send you errors.
My first attempt to deal with the messages was messy. I spent time looking at the various errors and came up with some common ground between them, like subject lines and the message bodies. I grabbed a PHP POP3 class which allows PHP to read and handle mailbox messages. Then I wrote regular expressions to try to find the email address in question, and then unsubscribe it from the mailing list. This processed a lot of messages but never handled them all…too many variations of error messages.
My latest method is to use something known as plussed email addresses. Some mail servers recognize an email address like email@example.com as email for firstname.lastname@example.org. Our server (sendmail) allows this. An application can use the “data” part to do something useful. Now I set the Errors-To to be something like email@example.com. I was able to ditch all the regular expression subject parsing and simply lookup the user id from the plussed data and unsubscribe them. This works great. (One note: I had to also set the Return-Path header to the same address to get this to work…not sure if some mail servers don’t like that + sign when it comes to bouncing messages, though it seems to work fine now)
Obviously I’m not handling the type of bounce error. So I’m probably unsubscribing some users who were just temporarily over their email quota or something. That means I have room for improvement. I’m not going to go back to full-scale regex parsing to find the error codes and stuff. Many bounces come with an attachment called “message/delivery-status” which often contains a standard format (I should probably have been using that file when I wrote the first implementation). My eventual plan is to use the error data in that file and implement some kind of scoring system which will allow for a few minor errors before I completely unsubscribe a user.
There are probably better ways out there to handle this stuff (email admins feel free to comment!), but I haven’t seen them, especially in PHP. I did come across a Perl CPAN module for handling bounces, but everything’s in PHP right now…I’d prefer to keep it that way.