On Friday, 13th March 2020 there was a public board meeting with an unexpectedly high attendance of nearly 30 people. It was noted that the virtual machine resources maxed out due to the load on Jitsi. While the result was disappointing, the strategy of the infra team had all along been to gradually increase the VM resources based on demand. There is no point in paying extra for something you don't use, after all. The VM resources were beefed up soon after the board meeting.
On Wednesday, 18th March 2020 we organised an ad hoc load testing session for Jitsi to see how the newly-increased resources affected the experience. The session included video streams and screensharing. Now the indication was that client software for some people was struggling with CPU load. Jitsi configuration was tweaked based on this: calls were made audio-only by default, for example.
Research into Jitsi's current development and known issues followed. Jitsi developers acknowledged that the web UI was heavy in some respects and said they were optimising it. A known issue was that Firefox users could have a negative effect on the bandwidth use of all participants in a call, when screensharing was used. We had not really considered this. All we knew since the beginning of TDF Jitsi deployment was that Firefox itself might not work reliably.
After running into a user report saying that Firefox was causing problems even with audio-only calls, we decided to do a simple experiment in a TDF team call (Tuesday, 7th April 2020). I joined the call as the only Firefox user for 5 minutes and then switched to Chromium. A person who regularly had substandard experience in Jitsi calls reported that the difference was night and day: with no Firefox users in the call, the quality was perfect.
A future with a browser monoculture in conference calls seemed terrifying, but thankfully both Jitsi and Firefox developers were working feverishly to make things better.
On Friday, 17th April 2020 a stable release was cut for Jitsi Meet with many Firefox-related improvements. It was deployed on TDF infra soon after. Based on the communication, the current nightly release of Firefox should work fine with the latest Jitsi Meet, while for stable Firefox (75) the only change was disabling a feature called simulcast (it allows the sending of multiple video streams with varying bandwidths/quality at once).
On Tuesday, 21st April 2020 I made another experiment in a TDF team call. I first joined with Firefox nightly for 5 minutes and then switched to stable Firefox for the rest of the meeting. The results were very promising: other participants observed no issues with Firefox nightly and only a single instance of audio cutting with stable Firefox during 30 minutes.
The current Firefox nightly represents what will become version 77. The stable release date for 77 is 2nd June 2020: https://wiki.mozilla.org/Release_Management/Calendar
All Jitsi Meet -related Firefox issues can be seen here: https://bugzilla.mozilla.org/buglist.cgi?status_whiteboard_type=substring&status_whiteboard=jitsi-meet&list_id=15210150
Bonus topic: during all the focus on Jitsi and its issues, it came up that it has rather poor accessibility support. Fortunately some people from the user community have already contributed patches that improve the situation during the past couple of weeks. Known a11y issues are tracked in this meta report: https://github.com/jitsi/jitsi-meet/issues/6090
Ilmari -- To unsubscribe e-mail to: website+unsubscribe@global.libreoffice.org Problems? https://www.libreoffice.org/get-help/mailing-lists/how-to-unsubscribe/ Posting guidelines + more: https://wiki.documentfoundation.org/Netiquette List archive: https://listarchives.libreoffice.org/global/website/ Privacy Policy: https://www.documentfoundation.org/privacy