Comment by wyc | Hacker News Reader

Comment by wyc | Hacker News Reader

Parent Full thread

Here's the "open stack" I've been trying to use for communications, veering towards web standards and open source:

Video with lots of participants: https://jitsi.org/

Video with a few participants: https://whereby.com/

Internal chat with tech-savvy folks: https://element.io/

I still use Slack and Zoom pretty often. People already know, love, and have installed these tools. It adds significant friction to expect everyone to use the tools you want, especially if they don't share certain values around technology, or if they're slow to adapt to new user interfaces.

I wish for a videoconferencing tool that:

- Built on open standards and runs fine on browers

- Is licensed Apache 2.0, MIT, or 3-Clause BSD

- Has slick native apps across GNU/Linux, MacOS, and Windows

- "Host yourself" server that scales horizontally, allowing me to throw more servers at this

- Supports screen sharing, recording, and transcription or integration with a 3rd party service

- Has an open chat API that allows you to hook up custom bots during meetings (Zoom is lacking this feature, but it's very important to me)

- Has other APIs build into the core workflows of conferencing, not implemented as an afterthought

- Supports an operating mode up to 100 simultaneous connections of HD video and audio

I would pay $30/mo per "host seat" for a SaaS implementation of the above.

If you ever get the itch to run this on AWS cloud, just don't. I ran Jitsi ( the whole system) on AWS on-demand EC2 for my Kid's class to collaborate on their projects during Covid lockdown (in Toronto) and it burnt a hole in my wallet. Not prophbitively expensive but still. So now I advise them to use Google Hangouts or Zoom. I often wonder how much Zoom and its competitors pay for bandwidth.

Very little, because AWS and GCP bandwidth costs are marked up by over 1000%.

Cloud services are sold the way drugs are.

The first hit is free. But after a tiny bit of, uh, scaling you're going to be ripping copper out of abandoned houses to make ends meet.

Do you happen to have the bill/invoice still? I'd love to see some real numbers, and try a bit of back-of-napkin comparison with digital ocean, scaleway, hetzner cloud - and in particular hetzner dedicated. The latter does cost a bit if you need 10gps uplink - but otherwise you could go far with a "second hand" auctioned server - free/included bandwidth (1gbps) and no setup fee. I suppose for the US market one might want to look at ovh as well.

Some competitors like Cisco opt to deploy their own AS (Webex is AS13445) optimized towards eyeball networks because capacity and latency to Google and Amazon can be tricky with some incumbents. So they generally have good control over their cost this way.

Yes, it's much more affordable to use a host like linode which can provide more reasonable rates.

So, a souped-up Matrix Server with Jitsi integration and a bunch of other plugins?

I'm not ready to offer that to you, but it's definitely on my roadmap for https://communick.com next year.

The thing that I'd suggest though: instead of coming with a laundry list of things that you'd like to see in a finished product, consider funding the development of projects that provide the pieces?

One of the pain points of Jitsi to me is the client-side muxing of media streams causing load shifted to laptops and phones, so not sure if Jitsi would work with this.

Does your service allow exporting data?

Yes, if you count that the services I am using (Mastodon, Matrix) provide methods to export the data directly from their clients.

If you want to get the data from your account (payment history, services you used, people you invited, etc) you can get that only via the API or semi-automatically if you request your data and I run my "export user data" back-office command.

Well, the thing is that you won't find a standalone product that does all this, because it's already difficult enough to get right some of those requirements on their own. These are just different verticals, so what you need is somebody who decides to integrate them into a cohesive final product. For that, the integrator will need to base off intermediate implementations of each vertical (or have huge resources to reinvent the respective wheels from scratch...)

Like for example OpenVidu [0] offers a platform over which standards-based video-conference products can be written. It takes care of WebRTC stuff, STUN/TURN/ICE shenanigans, user permissions, recordings, etc. so you can focus on e.g. building native apps with your desired chat solution. And even then, it in turn delegates the low-level WebRTC media transmission to Kurento, of which (disclaimer) I'm the current developer and maintainer.

[0]: https://openvidu.io/

We don't check all of your boxes right now, but we've been working on an API + app store for video chat apps. https://align.link/

Https://etherpad.org for real time Collabo doc editing.

Reply on news.ycombinator.com