So I set up a XMPP server in my homelab this weekend to see how the ecosystem has evolved since the last time I used it. Only took me a few hours.
It seems like it has pretty much everything we'd expect from IM these days? E2EE with forward secrecy, group chats, message archives, reactions, file uploads, audio messages, voice and video calling, location sharing, etc. etc. It's even federated!
Why are we sleeping on this tech? Is there some unmitigated downside I haven't discovered yet?
@jae I'll give it to you that onboarding could be kind of an issue, but #XMPP has come a really long way in that regards. From the server side of things (with mod_invites) just as from the client side with easy onboarding straight from the app without having to create an account from a web page or anything like that.
Onboarding from a recent (mid 2025) client is as easy as installing the app, open it, choose create account, pick one from the drop down list of server, provide a password. Done.
Yes, Delta is mainly made by a German company, so they have more or less the complete project in their hands. All clients are based on exactly the same library. That gives a nice "out-of-the-box" experience.
#Jabber/#XMPP is a huge and diverse ecosystem with sometimes conflicting interests or priorities. And it's an open standard with many implementations, which can result in incompatibilites.
I prefer the latter, because I see more advantages in that approach, but YMMV.
> Yes, Delta is mainly made by a German company, so they have more or less the complete project in their hands. All clients are based on exactly the same library. That gives a nice "out-of-the-box" experience.
yip, i work on the project the gerrman's have built the ecosystem to use the common core. it's fairly robust and has allowed me to write a tui for #deltachat
i've also created an xmpp <---> deltachat bridge in go that's almost ready to release (working on the omemo part still).
> #Jabber/#XMPP is a huge and diverse ecosystem with sometimes conflicting interests or priorities. And it's an open standard with many implementations, which can result in incompatibilites.
true, but the fact that we have this battle-tested protocol and you can build on is really nice.
Can't wait to try your XMPP-Delta bridge!
Is there a public repo somewhere?
My vague impression is:
Delta is focussed on private chats (individuals, small groups) and has its strength there.
#XMPP is deliberately a universal protocol for everything: private chats, team chats, public chats, social. In my company we even built an IoT product based on it!
Still, some #Jabber clients are focussed: #Conversations_im on private chats, #Movim on team chats and social features etc.
@debacle @gnemmi currently the repo is on my system. it'll open up once its working. the omemo is kicking my ass.
i think your take on deltachat is likely spot on. but we use email as a transport layer with some bolt ons for performance
there's not much call for interop with other platforms but i can't force it on people so i engineer bridges.
Idk. Never tried Delta, because it doesn't have a client my OS. Only from third party repos, which I try to avoid.
Trying the bridge will be challeging, because nobody in my bubble does use it so far 😉
@debacle @methuselah @gnemmi what operating system do you run?
> Trying the bridge will be challeging, because nobody in my bubble does use it so far
i've got people on both sides of #xmpp and #deltachat since both platforms i like a lot. ill figure it out soon i hope, but taking a step back to solve some ci issues i hope on an upstream repo
I'm on #Debian and I'm also co-maintainer of various #Jabber related packages (#Dino, #Gajim, some more).
Installing Delta on Debian is easy (they provide appimage, flatpak, or snap, AFAIK), but I *strongly* prefer Debian main, i.e. not third party repos. That's easy with both Jabber and Matrix, but not yet with Delta.
Situation might improve at some point, though:
Nice! I was thinking I'd do a deltachat/XMPP bridge one of these days, but one less thing to do for me. My little finger told me @daniel might maybe possibly maybe be maybe interested in doing proper e2ee with a deltachat bridge (proper e2ee=dedicated client support needed).
What did you make your bridge with? (Language? Lib?)
@nicoco @debacle @gnemmi @daniel the bridge is in go, it works bidirectional and for each contact or muc it will create a new deltachat group for the user and his actor (bridge account for xmpp). its designed for single users vs everyone sharing the same bridge endpoint.
its been a while since i got deep into xmpp spec but its been fun so far. missing piece is omemo on the xmpp actor account which is proving tough.
@jae @gnemmi yeah, I just ran into this actually. Told a friend that I'm using Conversations on Android. Then they tell me it's $7 on the play store. I was blindsided by that because I installed it from F-Droid, where it's free.
Installing F-Droid or paying the play store tax is high barrier of entry for non-technical folks to experiment with the tech.
Which is something I say with mixed feelings, because open source developers absolutely deserve to be compensated.
@smn @gnemmi i can appreciate that dilemma. for the newcomer, dropping 7-20$ on an app just to test often isn't reasonable or at least not a priority when signal/whatsapp/etc "just works".
but hey, many of us technical people are on xmpp (or dual like me on #deltachat) so we can still keep in touch.
btw what xmpp server are you running? i'm fleshing out an implementation based on an xmpp-go setup i found on github since i'm sort of obsessed with go. :-)
@jae @gnemmi I went with ejabberd, since that's what I used to run back in the day.
Very cool that acme support is built in these days, but I wish the docs gave better advice for how to expose port 80 to succeed at the HTTP-01 challenge. Fortunately I already had a reverse proxy listening on port 80 so setting that up for this project was simple. I do feel like a lot of novice or even intermediate sysadmins might get stuck here though.
Not thrilled about the plaintext password default either.
Speaking from a #Jabber/#XMPP fanperson perspective, the downsides are:
1. No good client with A/V calls for Windows. #Gajim by @gajim is on the MS store and is good, but doesn't have A/V calls, #Dino by @dino is also good, has A/V calls, but only has outdated Windows installers somewhere on the net. There is #Movim by @movim, but web clients are not everyones cup of tea. Problem can and will be solved, but atm. it is a drawback. Maybe #Kaidan by @kaidan will solve that.
2. Because of the long history of Jabber (since 1999!), and huge advancements in cryptography, encryption in XMPP has many options. So far, so good, but most clients are still on #OMEMO 1, which has some problems, and only few clients already have OMEMO 2 (#Kaidan by @kaidan, #Libervia by @Goffi, soon‽ also #Conversations_im by @daniel, #ConverseJS by @jcbrand and #JabberEl by @[email protected]). And there is no standard way to move decrypted message from an old device to a new one!
3. A/V calls typically work great for one-to-one calls, if network, TURN server etc. are working correctly. But multi-party calls are "not there yet", esp. for groups > ≈ 4 people. Both #Movim by @movim and #Libervia by @Goffi have it on their roadmap, and I'm optimistic, that it will work well later this year or the next.
That's the three downsides from my PoV, but Jabber is still my favourite chat.
@debacle @gnemmi @smn @movim Hey, SFU (big conferences) is not just on the roadmap, there is a working prototype for nearly 2 years now (based on the excellent Galène), with corresponding specs in XSF /inbox. Well was working actually, broken right now due to big changes in web frontend, but should be back soon.
Justin Fitzsimmons
Gonzalo Nemmi 
ᴏᴏᴍ-ᴋɪʟʟᴇʀ: 333
Debacle
methuselah
Nicoco
what the sigma
silverwizard
Ténno Seremél’
Goffi