Modern XMPP project discussion - 2024-04-28

Link Mauve, just a small mistake, its corrected now; it now gets up to: <failure xmlns='urn:ietf:params:xml:ns:xmpp-sasl'><not-authorized/></failure> for authenticating as anything@jabber.fr using SCRAM-SHA-1

arcanicanis, perfect, that’s to be expected, it means you can now go a bit after the very first <open/>. :)

Link Mauve, is it supposed to fail on SCRAM-SHA-1, or was the script that allows any password somehow only functional for PLAIN auth or similar?

There's no script that allows any password

Link Mauve was just saying that any password would reproduce the error

Oh, understood

Anyway, the point of the exercise is just to illustrate the point of how much more trivial it is to write a webclient from scratch (or perhaps a client in general), especially with how much is just built into the web browser now that doesn't need libraries (and in fact, possibly slower with some libraries that implement their crypto manually in plain JavaScript, versus the built-in features), that the foundation doesn't have to be anything complex

whereas my point on protocol complexity and such would be to challenge someone to try accomplish the same with Matrix (write a client that's able to send messages, with libraries)

and I hope to next get basic chat and WebRTC next, to illustrate how much web browsers have dumbed down a lot of the work of it for you, and how things like MS Teams, Discord, etc aren't novel and anyone could stand up something similarly on XMPP for signalling, or on whatever signalling transport (just as someone was trying to do it over ActivityPub even),

I just have apathy for people virtuing that they care about privacy, anonymity, etc, but then cling exclusively to Discord, when it's insane amounts of telemetry (virtually every action you do in the Discord client is logged and stored, along with location/device info on each event, plus contextual info to each type of event, stored indefinitely), when alternative solutions have already existed (and long before Discord even), just people don't bother touching them.

The only main complaint I sometimes get is about the cosmetics or menu structure of some clients, meanwhile the amount of actively-maintained graphical clients feels a bit narrow; e.g. the recommendations usually follow in: Conversations (or forks of it) for Android, Monal (which recently came back to life) for iOS/macOS, Gajim for desktop; Conversejs for a fully client-side webclient, Movim for a more mixed 'hosted' webclient; whereas each project looks mostly like a one-man show (e.g. developer of Conversations focuses on an Android email client for a while, and development of Conversations slows down a bit)

whereas to my understanding, XMPP seems relatively straightforward to work with (or at least now), and people could be contributing, but usually aren't; open source used to be about the point that anyone could jump in and contribute on mistakes (or possible areas of improvement) they find, versus snapping their fingers at developers expecting them to fulfill all their wishes for free

> the point of the exercise is just to illustrate the point of how much more trivial it is to write a webclient from scratch This hurts a bit :D

> > the point of the exercise is just to illustrate the point of how much more trivial it is to write a webclient from scratch > This hurts a bit :D I'm saying _to start_, I don't mean to trivialize your work at all

I'd say it's very accomplished for anyone to still be maintaining and developing an XMPP client after an amount of years, after listening to all the whining of people saying "how it should be done instead" when said people can't even be bothered to file an issue on GitHub or whatever for other matters,

Every time I follow a link to github, an angel goes to hell. 🙂

I'd strongly prefer if more projects could move to GitLab, Codeberg, or some self-hosted Gitea instance or similar.

I just mean more the point of people that complain, that can't even put in the most minimal of effort into anything actionable

e.g. I get complaints like "XMPP clients are ugly and outdated" and I usually ask "Okay, then how would you design it instead?" and never get an answer

Any test suites for Jingle, or any bots to aid with testing Jingle stream setup? Also, with TURN, is it the call initiator's TURN server that's used (if needed) for all participants, or does each peer use their own TURN server provided by their respective server?

arcanicanis: each side uses their own

I guess that complicates things then, given a lot of small servers usually don't have TURN deployed

These days? I would think they have, if they have a modern client. Otherwise they can't call...