I personally found investing a little time to learn docker-compose very worthwhile. I already run NC, mariadb, coturn, redis, and Collabora in Docker. I plan to try adding Jitsi soon, possibly this weekend.
Pre-tested docker-compose files can make setup very easy.
Also, take the concerns surrounding Docker and root with a grain of salt. Even on ârootlessâ systems there are still many processes running with root privileges, and Nextcloudâs Docker version runs as www-data inside the container.
founder of nextcloud⊠yes. founder of talk? ummm. no. as i have learned that talk/spreedme seems to be an external software which got included to NC as an app. so founder-frank might like the idea but it seems to be out of his reach if talk/spreedme would do soâŠ
I played a lot on Jitsi those days and i think the Jitsi app « may » be hard to implement.
The way Jitsi implement security about room creation is akward. TL;DR everyone can access room creation and enter a room, but itâs only after that you to input credentials for using the room.
I hope dev can modify this behavior and find something simple to use and share.
I will now look on Big Blue Button to see what the software can produce.
In Matrix the original Jitsi UI is shown in a widget (like a small browser window) and the client sets a random room name. Itâs not a deep integration at all (at least not yet).
The newest Riot (Web) Client has a config option to set your own jitsi server. Tried that just yesterday, itâs very simple to set up.
If your Jitsi server is password secured you have to click inside the tiny widget and enter your credentials.
I think other systems only use session-ids and thereby a hidden room name. Perhaps the jitsi-room-name can put from url to post-parameter and do not show in the browser and app. I think the jitsi devs can easy program a second authentication with (hidden) session ids.
I just wrote a complete streaming Nextcloud Talk API proxy and now pretty advanced Jitsi chatbot with full password/authentication support:
I created a short introduction video to the streaming API/bot framework here:
If you know German, take a look at the following usage example video:
I now use this setup on a 300+ concurrent user Nextcloud instance and have received very positive UX & performance feedback (people are used to Chatbots from Slack and Discord), so this might be a pretty good solution until an actual integration has been created.
@jakob1 Theoretically yes, but because of Nextcloudâs IMHO âretroâ approach of how they do apps (which is really not compatible with most modern/âcloud nativeâ extension ideas like i.e. Kubernetes does it with APIs/gRPC) it isnât that simple. If one where up to the task though, it would of course be possible to simply call Docker using an app and then controlling the chatbot with it, although I made an oath to never tough PHP again
i am also not impressed from Talk. If i have more than 3 or 4 Participants it is extrem laggy and the Bandwith needed on each client ist high. And i see only 5 to 6 ⊠everone else is out of the Window ? So for self hosting it is not usable at all for me.
I played also around jitsi and setup my own Jitsi Server Instance ⊠well performed with 10 Participants without any problems and by the way i see all Participants without problem, can choose which one should be in big or spreed all over the Window - great. Video and Sound quality was good.
So i am also interessted in an nextcloud integration as an optional Videosystem to Nextcloud Talk. If jitis is still working in the next days without problems i drop nextcloud talk because it makes no benefit for me.
All you need to get Jitsi like performance is to write an open source HPB signaling server, been working on one in my spare time. Basically just a web api and a Janus server put together.
the Jitsi app « may » be hard to implement.
of the auth/login method, i know that it can be modded to fit the user base of Nextcloud.