Tuesday, January 14, 2014

Modifying mail passing through Zimbra

I had a request to attach to (almost) each mail message that passes through the Zimbra mail server an image. Basically, what the owner wanted is that there is an image with advertisement in the mail that is sent by internal users. Additional requirements were:
  1. Image should be added only once!
  2. In case there is new image, and there is already on in the mail, the old one should be replaced!
  3. Image has to be at exact spot within a message.
There were some additional requirements from my POV:
  1. Not every mail should have image attached, e.g. automatically generated internal messages!
  2. I should be careful about impact on the performance.
  3. Mali messages that are not modified should not have any noticeable marks about image that isn't added (this one will be more clear later).
  4. The solution should allow to define only certain senders to have mails modified.
  5. It has to have a DRY RUN mode so that it is easily disabled.
I immediately knew that there is no way to place image somewhere on the Internet and put link into the mail message. Although the most elegant solution, the problem is that all mail clients don't show images by default, and that's it. So, image has to be within mail message itself. A bit of research showed up a potential solution. Namely, to embed image data within IMG tag itself, and mail messages are already altered by adding disclaimer which is HTML and a perfect place to add that IMG tag, so why not reuse disclaimer for the same purpose? In favor of that solution was also my intention not to add new scripts into the mail processing chain because of fear that it might impact performance. 

Unfortunately, that solution didn't work. The most important shortcoming is that Outlook and GMail don't handle IMG tag with embedded image data in it, in other words, the image isn't shown. To solve that, image has to be embedded as a MIME part within mail. Additionally, the other requirements aren't easy to achieve, especially, the one to replace old image with a new one. So, in the end I had to resort to writing scripts.

I already wrote about how I managed to solve more complex requirements for a disclaimer than the functionality of Zimbra allows. So, it was natural place for me to add that additional processing there. I called the script to add image altermail.py and now altermime script has the following form:
#!/bin/bash
grep "DISCLAIMER:" ${1#--input=} > /dev/null 2>&1
if [ ! "$?" = 0 ]; then
/opt/zimbra/altermime-0.3.10/bin/altermime-bin "$@"
fi
#echo "`date +%Y%m%d%H%M%S` $@" >> /tmp/altermime-args
/opt/zimbra/altermime-0.3.10/bin/altermail.py "$@" >> /tmp/altermail.log 2>&1
I'm calling altermial.py after altermime because the image placeholder is within disclaimer! Also, I removed exec keyword before altermime-bin call so that altermail.py is finished.

Additionally, note that altermail.py accepts the same arguments as altermime! This is in order to simplify things a bit.

Obviously, I choose Python as a programming language of my choice. I could write all that in Perl too, but since I'm lately working a lot more with Python, Python was the way to go. Both languages have very good support for mail processing (MIME messages in particular).

The script is on the GitHub and you can fetch it there.

How the script works

First of, script doesn't work for mail messages that aren't MIME. So, after loading a message the first check if it is a multipart message. If not, then script just exits.

Next, white and black lists are checked.

There are two passess over the mail message. In the first pass, it searches through the mail message to see if there is already image attached. If so, then it additionally checks if it is an older version of the image. If it is, it replaces the image, but in both cases it doesn't do anything more and finishes execution.

The second pass is done when there is no image in the mail message and it has to be added. Now, when adding the image it has to be added with HTML as html/related so that they are both shown. If you add it as html/alternative, then only one of them will be shown!

All the configuration options are embedded within script itself. I chose not to have configuration file to reduce number of disk accesses, which is already very high (a lot of modules are necessary).


No comments:

About Me

scientist, consultant, security specialist, networking guy, system administrator, philosopher ;)