[Mailman-Developers] Mailman 3 status

[Stephen J. Turnbull]

Hi, Mailman Users!

Executive summary: In my opinion, Mailman 3 is ready for use in
production, but still requires skills and system accesses over and
above those required by Mailman 2.  Both in principle and in practice,
questions about Mailman 3 should be directed to the Mailman Developers
list (preferred) or the #mailman channel on Freenode IRC (where there
is an intermittent presence of relevant developers).

Note to Mailman Developers: Reply-To set to Mailman Developers.  If
I've made any misstatements that Mailman Users needs to know about,
make sure your post is addressed appropriately to Mailman-Users.

What Mailman 3 has
------------------

Mailman 3 has a pretty complete reimplementation of Mailman 2
functionality with a few exceptions.  However, the implementation has
been radically redesigned, dividing responsibility for Mailman
services into

- Mailman core: managing user databases and message redistribution

- Mailman client: the Python API for interacting with the core, which
  actually communicates via HTTP using a RESTful protocol

- Postorius: the web interface for managing sites (including virtual
  hosts), lists, and subscriptions

- HyperKitty: the list archive manager

For the exceptions, please see the release announcement or for details
ask on Mailman Developers.

Note: we consider the functionality provided by the modules listed
above to be quite robust already.  We're not ready to say you should
migrate your 24x7 security announcement list of 100,000 subscribers to
Mailman 3 yet, but Mailman 3 is already serving many discussion lists
successfully, and the number is increasing at a satisfactory pace.

What Mailman 3 lacks so far
---------------------------

- The "topics" feature.

- Ports of most Mailman 2 add-ons and commonly-used patches.

- Performance measurements.  There are theoretical reasons to believe
  that under certain circumstances a "large" Mailman 3 under "heavy"
  use *might* suffer bottlenecks, but we just don't know yet.

- A Mailman 2 to Mailman 3 migration story.  We know what to do, and
  we can tell you how to do it.  It's not hard for simple Mailman 2
  installations with a few discussion lists, but it's not automatic.

- A "simple" installation story.  Because Mailman is now divided into
  three separate projects ("client" is closely bound to "core") with
  separate development groups and release cycles, installation has
  become more complex.  It is still recommended that users install
  recent versions with the latest bugfixes (and for special needs,
  even more recent versions with feature improvements) from the VCS
  repository on GitLab.

  We do have the "Mailman Bundler" script which automates the process,
  but it is as yet lightly tested, and mostly in use by experienced
  system administrators rather than "mom and pop" list owners.  Some
  users have experienced a bit of frustration, although so far we've
  been able to provide workarounds "on the spot", and quickly improve
  the scripts.  Often these improvements are provided by users, which
  indicates the level of experience of typical testers.

  As yet there are no packages for Mailman 3 in "stable" or "long term
  support" releases of OS distros, so that avenue is also closed to
  folks who don't want to be "guinea pigs" (ie, early stage beta
  testers for installation and migration).

- Hosting service (cPanel etc) integration.

- User-tested documentation.  Again, most of our beta testers are
  experienced system admins.  Very likely they are reading between the
  lines for details that people whose administration experience is
  limited to Mailman lists, probably on a cPanel system, won't get.

- A support system on Mailman-Users.  I'm sure you all are well-aware
  of how spoiled you are to get support from Mark Sapiro, with
  occasional contributions from users and other developers.  Mark is
  still the lead developer on Mailman 2, and has not yet gotten a lot
  of experience with Mailman 3.  We do not have an equivalent person
  for Mailman 3 yet, although there are several developers any of whom
  might grow into that role.[1]  You still really need to contact the
  developers directly:

  - Mailman Developers <mailman-developers at python.org>

  - #mailman channel on Freenode IRC.

Note: the "lacks" column will be mostly a barrier to non-system-
admins, but of course may be a strong "not quite yet" signal even to
experienced admins.

What Mailman 3 will have after GSoC
-----------------------------------

- An experimental implementation of Dynamic Sublists, which aims at a
  substantial improvement in functionality over the "topics" feature
  of Mailman 2.

- An experimental implementation of a client based on Node.js,
  allowing web frameworks based on Node.js to talk directly to Mailman
  3.

- One or more primitive Mailman 2 to Mailman 3 migration stories.

Note: Integration of these features will probably happen on VCS
branches in September to October, and will be considered ready for
beta use in version 3.1.

What Mailman 3 plans for users in version 3.1
---------------------------------------------

- A simple installation story, probably based on the current "Mailman
  Bundler".

- A simple 2to3 migration story.  Since we don't have any scripts yet,
  this will still be considered experimental/direct user suppport
  needed for Mailman 3.1.  "Our experienced staff will be on call."
  (At least on the Mailman Developers list.  I personally will push
  for attention to Mailman Users as well.  Finally, there has been a
  suggestion of "mailman3-users", also -- watch for details in the
  release announcement :-).

- Improved list-owner-oriented documentation.

A final note
------------

The information provided above is accurate to the best of my ability,
but it has not been checked by the responsible developers.  It is
provided in hope of may be of use to those considering installation of
Mailman 3.  If you're still on the fence after reading this post,
please do get more accurate information from the responsible
developers on the Mailman Developers list.

Footnotes: 
[1]  I really can't speak with authority here; this is my impression
though.