I've been playing around with self-hosting Jitsi Meet for online chats. And I've been playing around with the podcasting studios at a local library. So I've also been playing with recording remote interviews using both.
The last of these didn't go well. It turns out that accessing the Jitsi Meet instance on my server (or indeed any old server) is no problem at all. Works a treat. Unless you're behind a firewall that blocks User Datagram Protocol (UDP). It turned out that remote video and audio didn't come through (although text chat worked), and when I had two laptops (one just to record the session) connected simultaneously on my side, the connection dropped entirely every minute or two, requiring logging in again every time.
Internet video and audio streaming typically requires using the UDP protocol, as it is low latency, as opposed to Transmission Control Protocol (TCP) used by Internet services that require reliable transmission of data. Reliability requires checking that you got all the data you asked for, uncorrupted, and in the order you asked for it, and also requires asking for it again (and again, and again…) whenever something goes wrong, and potentially waiting a substantial fraction of a second for every missing or screwed up data packet. So reliability increases latency considerably.
Now recorded video can use TCP and work around that problem by buffering the video ahead of playback. This isn't a viable solution for real time conversations (because you haven't said what you said until you said it), so in practice internet teleconferencing involves using UDP and just tolerating any lost/mangled data. This is why Zoom/Teams/whatever, despite the massive throughput of their data centres, still manages to be glitchy as hell on occasion. (While YouTube and Netflix are high-res and flawless.)
The problem is that UDP is also used for all sorts of other applications that corporate networks regard as illegitimate, such as gaming or censorship evasion, and is also often used in malicious attacks. So corporate firewalls often simply block UDP access overall.
"But Katy", I hear you protest, "It's not the business of libraries to prohibit play, and certainly not to facilitate censorship! And who would maliciously attack a library?"
I quite agree, but the first priority of IT departments is arse-covering, and "Everybody else does it!" has been the default reason for pretty much everything IT departments have done since time immemorial.
"But why then," you argue, "Does Zoom and Microsoft Teams work perfectly well at my office and at the library?"
My, you're stroppy today! Is everything okay?
This is the fun part. Your office and library firewall will whitelist Zoom, and Teams, and countless other big tech services, because these are either services they have explicitly invited into their organisation as a partner, or they are just so pervasive that they can't deny access to their users. These companies have the heft to demand that basically every other company on the planet reconfigure their network security to accommodate them.
So for every potential TCP or UDP connection to every public-facing server of the tech giants, network administrators at the user end are in a constant scramble to keep their long whitelists up to date. Because of course, servers are being added and removed all the time. Microsoft says it updates the list of it's TCP/UDP addresses and ports monthly.
Sources:
- Prepare your organization's network for Microsoft Teams
- Managing Microsoft 365 endpoints
- Zoom network firewall or proxy server settings
I read somewhere today (on a discussion board; dated; not trustworthy; seems plausible, though) that, according to data from Google's not-at-all-creepy omniscient surveillance capabilities, a bit under 10% of Web traffic goes through a corporate firewall.
This is a part of the dynamics that are breaking the Internet. Not only is this a huge barrier to market entry to a new online service; one that increases the market power of existing players, but it kills stone dead the idea of the Internet as a level playing field where the infrastructure is dumb and undiscriminating, and all the data and the services that use it are owned and controlled by users.
Worse, even a quite large corporation likely doesn't want to pay someone just to keep on top of network security that operates on the sledgehammer principle of block everything not explicitly permitted. However, if they don't, now-critical third-party services will regularly break. Which encourages outsourcing this job to other tech giants like Cisco, who than use it as a "land and expand" opportunity to colonise the rest of the IT department.
Now is 10% of your audience (or whatever it is in reality) a showstopper for my purposes? I'm not sure, and in principle at least there appears to be a sub-optimal workaround for UDP blocking that looks like a bit of a pain but worth evaluation. For conversations with people you know are not behind somebody else's firewall (which they could be if at work, a library, cafe, pub, maybe a mobile data service, etc.) it's fine. (I could even see if my library will add my server to their whitelist, assuming that's something they control.) For online events open to the general public, less fine.
Investigations continue.