Hi,
Riseup plans to stop their XMPP service in August 2021. We're using
that service for a few public chat rooms, most notably "tails-dev" and
"tails".
Proposal: migrate these 2 chat rooms to another friendly XMPP server.
We're considering 2 hosting options for now (systemli.org and
esiliati.org). Hosting suggestions are welcome, in case none of these
2 options works out.
boyska and I have been looking into this problem space: our needs and
options. A few other people have participated in discussions on this
topic in various places (one could say that instant messaging is a hot
topic these days). Thanks!
Here's a summary:
https://gitlab.tails.boum.org/tails/blueprints/-/wikis/Migrate_from_xmpp.riseup.net/
I'd like to hear your thoughts about this topic by July 11.
Before you comment, please read the kind-of-FAQ that follows.
Thanks in advance!
Why keep a public chat room at all?
===================================
A few months ago, in a survey run on this mailing list, a number of
reasons were raised in favor of keeping a tails-dev public chat room.
sajolida played the devil's advocate about some of these reasons,
then the discussion stopped. My understanding is that we have a rough
consensus in favor of keeping such a chat room, at least for now.
What about the "tails" chat room?
=================================
I've not seen anyone advocating in favor of keeping the "tails"
chatroom, but nobody explicitly proposed to ditch it either. It does
not look very useful but if the solution we pick for "tails-dev"
cheaply works for "tails" too, the latter can follow along.
Why not IRC?
============
The main reason is that to get any kind of backlog on IRC, you need
a nerdy setup. I believe we should optimize the Tails contributor
experience towards younger generations who have never heard of an IRC
bouncer, towards people who don't easily have access to
SSH+{screen,tmux}, and towards people who are not at ease yet with
command-line interfaces.
Another, weaker reason, is that registering through Tor can be painful
because of reCaptchas. And we'll probably want to require a registered
account to join our chat rooms at some point (that's the only way to
reasonably deal with abuse), so this does matter.
Why not a more modern chat protocol?
====================================
Compared to IRC, modern chat protocols such as Matrix provide a nicer
UX and a feature set that's closer to the expectations of users in
this age. In theory, XMPP achieves the same; but it depends so much on
what optional protocol extensions are implemented by the clients &
servers, that in practice the feature set one can count on is much
smaller. So it's very tempting to migrate to a more modern chat
protocols. For example, GNOME and Mozilla are in the process of
migrating to Matrix.
However, I prefer not to open this can of worms right now:
- The timeline of migrating out of Riseup's XMPP server is pretty tight.
- Many of our core people are busy in July.
- Instant messaging is a hot topic. Strong opinions are a dime a dozen.
- We did not specify our needs and desires yet, regarding what sort of
extra features we would like. Heck, many of us have no idea
- Friendly hosting options for modern chat servers are not very
numerous yet. E.g. systemli.org runs a *beta* Matrix server.
Note that this could change: I expect friendly XMPP hosting to
slowly (?) be replaced by this sort of things.
- If we introduce another protocol, it would be nice to also migrate
our private, internal chat rooms to it, so that contributors
only need 1 chat client, and possibly we can drop one piece
of infra we're maintaining. We would need to discuss
(probably elsewhere) the corresponding hosting requirements.
I don't think our group has the capacity to go through all this within
a month.
I'd like us to come back to it later, possibly in a year or 2.
Why not bridging protocols?
===========================
I don't think we should bridge chat protocols: the impedance mismatch
is so big between these protocols, and their feature sets are so
different, that experience from other projects, most notably GNOME,
shows that bridging
1. is difficult and costly to set up in a way that works better than
a proof-of-concept: it took GNOME months working with the Matrix
people to have somewhat vaguely good enough
2. creates a worse UX for all users, regardless of the protocol they
use to connect, than using a single protocol:
https://discourse.gnome.org/t/irc-matrix-and-thanks-for-all-the-kicks/6482
Cheers!
