search

datalad-datasets / ohbm2020-posters

Centralized registry of Jitsi rooms and pdfs for OHBM 2020 posters.
https://datalad-datasets.github.io/ohbm2020-posters
32 stars 84 forks source link

Bunch of wrong attendance numbers #154

Closed yarikoptic closed 4 years ago

yarikoptic commented 4 years ago

I think nobody is in any room, but we have non 0 counts. Hard reset to 0?

soichih commented 4 years ago

They are non-0 because we are receiving keepalive from various users still. If we reset them to 0, it will just go right back again. We also have code that automatically purges stale client IDs. So if client fails to send the jitclose, it should be removed within 60 seconds.

I am hoping that.. with PR #150, once everyone refreshes their pages, we should no longer see those phantom counts - caused by someone having a tag / page opened with closed jitsi sessions. But the time will tell.

Oh.. and I need to have @r03ert0 make similar update on brainweb version.. the counts might be coming from theirs if they haven't updated it.

soichih commented 4 years ago

Might be caused by #153

r03ert0 commented 4 years ago

done!

soichih commented 4 years ago

@r03ert0

I just submitted this PR

https://github.com/datalad-datasets/ohbm2020-posters/pulls

I've made quite a bit of changes to room.html.. but basically, you can now send the actual count of participants as reported by the jitsi API (api.getNumberOfParticipants()) and if this is set, the server side will trust it as the actual participant count (ignore the unique IP list). This allows us to accurate report the number of participants including people who didn't join through our web Apps.

Another important update is that, the room.html will now only start sending keepalive after jitsi client is actually opened (videoConferenceJoined event). On mobile phone, user is given a chance to choose between web version and the App version. If the user sits there without making a decision, the count should not increase, and I believe this was happening on some cases.

I've also made various Vue related updates.. which makes it harder to read the code if you are not familiar with Vue.. but I apologize if that is the case!

yarikoptic commented 4 years ago

I think @soichih has addressed many possible use cases and numbers became much more robust etc. Closing