That said, the move will likely be to the next highly funded product in that case, not back to TS/Mumble.
We only pay $75/month, for two levels of server boosting. It's so underpriced for the corporate usecase, it's practically criminal. We'd probably be willing to pay four or five times as much, especially if it would allow us to host video/screen-shares with more people.
I appreciate that Discord is gamer-focused branding, but their inability to launch more or less the same product under a professional brand is astounding. They're leaving huge sums of money on the table. For example, being able to run a public Discord instance for customer support, with individual rooms per customer, and customer screen sharing and get anybody in the company to leave their team office on the private instance and join the support call in two clicks is mind-blowing.
b) Granted that you accept the risk of your vendor recording your communications, the odds of those communications being recorded are much, much lower for voice communications than they are for text communications, for the sheer cost of storing voice and video data. Running the audio through speech-to-text before discarding the audio may be a threat, but not for users whose office lingua franca is in a language not supported by contemporary speech-to-text tools, let alone automatically deciphering which language the users in a given voice channel are speaking without it being defined ahead of time.
In short, as long as you're not buying communication services from a competitor, you're probably fine.
Audio maybe - but stored indefinitly? No.
I hope they monetize in that direction rather than ads (Discord seems to still be pretty reliant on VC funding, which makes sense to me since Nitro is pretty cheap and not particularly necessary to use the product) — it seems less soul-eating. They've been dialing back the gamer branding at least, which could help reach a broader audience.
I haven't read into this but it seems like they're doing peer-to-peer for audio or the server load is INCREDIBLY efficient, given this.
This is observable to still be the case as an end user. When the discord servers are having issues, people's voices become inaudible. This can be resolved by moving the server to a different, less loaded region, even if that region should have worse connectivity to the users in the voice chat. For example, my raid group often have to bounce our server from Europe to the US to workaround reliability issues on the EU servers.
> link describe a P2P WebRTC implementation
what am i missing?
edit: found it. they use a SFU forwarder. That's the cheap version of going full centralization (MCU) benefits. they probably have a better deal on bandwidth than CPU to justify that, or all their use cases work fine with only one person transmitting at once, which gives you the same benefit of a MCU with less cpu usage...
They also do gif transcoding and a bunch of other things that do take actual compute resources that cost money.