Mail filtering for Apache REEF

[Markus Weimer <ma...@weimo.de>]

The REEF developer mailing list gets an enormous amount of messages,
mainly from three sources:

  1. Users and developers send email to the list.
  2. JIRA sends its updates to the list whenever an issue is created,
     updated or commented on.
  3. GitHub sends notifications about pull requests there.

On a busy day, this leads to 100 or more emails to be processed, a task
that can be eased greatly by using mail filters. In this document, I
will describe my setup in the hopes it helps other devs:

Mark mail from yourself as read
-------------------------------
Automated emails are triggered also for your own actions, e.g.
commenting on a JIRA issue. Hence, you will get an email from yourself
with content you just wrote. As that is a waste of attention, you should
mark those messages as read immediately. This can be tricky, as you have
many identities on the various systems. For me, those are:

  * `weimer@apache.org`
  * `Markus Weimer (JIRA)`

Depending on your setup, there can be many more. It is convenient to
package them all into one rule, if your email system allows that.

Separate automated mail from human generated email
--------------------------------------------------
I find it convenient to separate the mails from different origins into
different folders or labels (GMail):

  * `REEF JIRA` containts all emails sent to the *list*
    `dev.reef.incubator.apache.org` whose *subject* contains `jira`
  * `REEF GitHub` contains all emails sent to the *list*
    `incubator-reef.apache.github.com`.
  * `REEF Dev` contains all emails sent to the *list*
    `dev.reef.incubator.apache.org` whose *subject* does not contain
    `jira`

Not all email systems can filter by the `list` email header. In those
cases, you have to replace the list-rule with corresponding from: and
to: rules, which are somewhat imperfect and will leak messages to your
inbox.

Use an email system that generates proper mail headers for threads
------------------------------------------------------------------
Many of our discussions happen in threads with long chains of back and
forth. That is reflected in the mailing list archives. This threading
relies on the mailing system of all participants to connect replies to
the original message via IDs in the mail header. Sadly, many commercial
email systems don't do that. Mozilla Thunderbird is an open source cross
platform email client that does. It also visualizes the threading
structure and can be configured to only show threads with
unread messages.