NixCon 2020 Online Edition ideas

This is a continuation of NixCon 2020 status update

We have the idea of organizing NixCon online instead due to COVID. But how should it look like? It’s not super clear to me how the conference would translate when happening online.

  • Should we organize talks, interviews, …?
  • Should we have hack day tracks with a room per topic?
  • Are there any existing examples of online conferences out there?

The goal of this thread is to clarify that vision because without that the other discussions are quite pointless. What is a format that you can imagine that would interest you?

In that regard, please avoid talking about specific tools unless framed under the light of possibilities that they can bring. <3

13 Likes

We have the idea of organizing NixCon online instead due to COVID. But how should it look like? It’s not super clear to me how the conference would translate when happening online.

  • Are there any existing examples of online conferences out there?

I am not saying my datapoints are the best or the most relevant, but here are the observations where I can claim eyewitness status.

Recent (yesterday!) example of a technical (Lisp) conference converted to online.

ELS-2020, 27–28 April 2020.

Description by an attendee (me):

  • For each accepted paper the authors were asked to provide a prerecorded talks
  • Talks were collected and streamed via an event channel on Twitch.tv
  • Many talks were video of slides + voiceover; some decided to do slides+face, side by side or cutting out a rectangle in the corner; the one speaker who actually green-screened impressed everyone
  • During the entire event (talks and breaks), the stream chat was actively used
    ** Before each talk the organisers asked the authors to say someting in the chat, so people could @-mention them for questions; during the talk there was a general discussion of the talk context, in addition to questions to authors (obviously, there were also reminders from organisers to not stray too far off-topic…)
    ** During the breaks the stream chat was a general conversation — pretty lively most of the time
  • As a weak replacement for a conference group photo, people were encouraged to photo themselves with the conference stream visible in the background; these were stitched together
  • There was an attempt to do some non-text chat between participants; an address on a Jitsi server of one of the participants existed; I think around 10 people we gave up and switched off the webcams, peak membership seemed pretty below 20

Observed impressions

  • Some people were happy that they can participate in something after not being able to travel to previous ELS editions
  • Definitely some focus of getting together with a wider discussion than normally easily observed in the open
  • With main global chat on Twitch.tv, there was no clean and common way to quickly break out a multi-person discussion on a narrow topic; that was seemingly accepted as «would be nice to have, but we can’t have everything and on a short notice»
7 Likes

Thanks for sharing your experience, that’s super helpful and brings some ideas.

That’s one of the nicest aspects of conferences in my opinion; having random encounters that happen by overhearing somebody say something. But also seems hard to reproduce online.

2 Likes

My previous post was «reporting», and in contrast please consider this one pure opinion.

A summary of what I observed is «There are many options for showing things, voice narration is appreciated by many, few complain about not seeing faces in real time». Note that scaling voice, especially if you ask everyone to mute-unless-speaking, is unbelievably simpler than multi-way video.

Another guess is «whatever you do, audience questions will end up being a text chat». Colleagues here in the university share that they would prefer the students to speak instead of asking via chat but have no idea how to achieve that!

That’s one of the nicest aspects of conferences in my opinion; having random encounters that happen by overhearing somebody say something. But also seems hard to reproduce online.

  1. I completely agree.

  2. Looking around at how various tools are used, I believe this can be achieved, if it is specifically targeted and there is enough time.

Preparations would need the following to play out, and some of the steps needs wallclock/calendar time as well as ability to reach a wider community efficiently. I actually think Nix* community has better ability to have wide opinion polls than Lisp community, and planning 4+ months ahead directly an online event leaves us some luxuries not available when mid-February the situation changes for an end-of-April conference (with significant amount of work done for physical organisation which probably also needed to be rolled back…)

First, we would need to decide on preferences.

  • What number/size of voice minigroups do we want? Any willingness to spend any kind of resources on anything multi-participant video?
  • Whatever we do, voice requires more precise human synchronisation and more «let’s all gather» overhead than text chat, so I guess the first thing to have is first-class procedure for text-chat-based random encounters?
  • Here we have the nice question of priorities: do we prefer these to be logged or unlogged?
  • What features we want? I guess some form of webchat, are there any other feature preferences?
  • Do we want any kind of integration of chat and presentations, or can these be two separate platforms, or do we just want them to agree to being iframes on the event page?
  • We probably need some kind of accounts for sending messages, what kind of accounts do we want to integrate with? Say, existing Freenode accounts? Existing GitHub accounts?
  • There is for sure quite a bit more stuff…

Note that preference collection requires calendar time, and community reach, and it cannot be done faster by simply working harder.

On the fly creation of spontaneous breakout rooms is not that rare; it happenned all the time in Nix* IRC before a more or less accepted decision was made to have a bounded number of rooms with clear-ish topics and a designated official general offtopic room. Same can happen in more or less most chat solutions.

Twitch chat is annoying in the following way, by the way: small failures require a reload to get the stream reception working again, and of course this loses the recent chat history.

Second, with preferences clear, we can actually evaluate the options and pick a few which provide the best support for the prioritised features.

Third, I guess someone would play with test deployments, and then there will be proposed procedures for no-experience users, like we now have the default step-by-step-ish recommended procedure for backporting commits. The idea is not just to describe how to use the platform, but also to establish a protocol that allows all that room-switching to happen with less friction and better understanding of other participants’ intentions.

Finally, there is spreading the instructions, and maybe final bikeshedding between a few competing drafts.

This step is also un-acceleratable, as we want everyone to hear about the proposed protocol and have time to read it!

1 Like

Thanks again for sharing your idea, it gives me quite a bit of food for thoughts.

Something that comes to mind is that I often find it annoying when people turn off their cameras during a video call because then I don’t get visual feedback anymore. We probably want to set some etiquettes on how to engage as you said.

1 Like

Something that comes to mind is that I often find it annoying when people turn off their cameras during a video call because then I don’t get visual feedback anymore. We probably want to set some etiquettes on how to engage as you said.

What size of videocalls?

Experience with small videocalls at the chair:

  • we have «video coffee break» with explicit goal «not forgetting each other’s faces» — I think we have not ever reached 9 videofeeds, and typical number is at most 6
  • for research discussions — I think currently there is a group of three of us coauthors where camera is used if and only if the senior coauthor participates in the call (he switches the camera on, then the rest follows suit), otherwise we only use voice; in a group of four of us coauthors interesecting with the first group of three only by me, we do not seem to even consider the idea of switching on webcams
  • other people in the chair translate the webcams of the main lecturer and TAs sometimes during the course meetings

So I have an impression that seeing the faces has significantly different value for different people.

For a large event one needs to budget the resources, of course.

  • A talk, multiple tens of listeners — no problem translating the speaker’s face, of course, but I have a vague feeling of impeding doom when I think of all the listeners having webcams on.

Among other things, I imagine a wall of 50 webcam feeds and I am not sure how usable it would be.

  • Small break-out discussions — it makes sense to have videocalls there, but then it is a good idea to estimate how many of them at once will happen. Maybe not too many, though.

  • But text will probably dominate anyway

You probably want to have the talks prerecorded in some way that fits the chosen platform (Nokia. Connecting people — Caterpillar. Disconnecting people), and then there is no feedback mid-talk, so it is a good idea to reintroduce it but not clash with the existing audio. So, chat.

(My belief is also that for a quick exchange it is faster to create a new room and post its name than to negotiate a moment when everyone can join and audio chat — meaning exclusive coordinated time; I am not saying I am sure)

  • Both for videocalls and for breakout chats, fire drills are useful. In the videocall case, to find out who hates what for which reason and what hates whom in which way, and to convince people how much load is perfectly fine. In the text chat case, to see how the proposed interaction flows interact in a turbulent high-flow chat.

I hate to be “that guy”, but Mumble has been able to fullfill all requirements that were listed above

  • easy to host
  • scales to thousands of people
  • stable connections
  • ad-hoc breakout rooms
  • arbitrary complex linking of channels

for years, and in spades.

No doubt live webcam feeds are nice, but the current technical state of the world is just not at a point yet, where that’s an option for everybody

  • bandwidth issues in 3rd-world internet countries (like Germany :slight_smile: )
  • can’t expect everybody in a community like NixOS to install a proprietary client with lots of questionable “features” (sorry zoom-us) on their systems
  • F/OSS solutions like Jitsi or BBB are not there yet (and there’s an open question of hosting costs for such a service)

I suggest we would combine

  • pre-recorded talks with live Q&A sessions with the speakers, in a more traditional “tracked” conference
  • breakout channels for workgroups/hackathon (this part can be structured like an unconference)
  • Random ad-hoc channels for open discussion that can set the current discussion topics in the channel name/description
  • Some way for people to do self-organized sessions (see the 36c3 page to get a feeling how that can look)
8 Likes

I hate to be “that guy”, but Mumble has been able to fullfill all requirements that were listed above

(looked at Mumble) Yes, definitely as good as one could ever want a client-based solution to be.

  • F/OSS solutions like Jitsi or BBB are not there yet (and there’s an open question of hosting costs for such a service)

I object: BBB is nearly perfect! for 1–5 instructors with at least a minimal BBB experience teaching a classroom of 10–100 students in a way remotely similar to some existing offline classroom workflow. and in a situation where using the browser as a client is desirable. Which is its goal, and which creates some friction when using it for anything else.

I have no experience with Mumble but I trust yours, and the general features/layout seem very good for our use case, I agree.


I suggest we would combine

  • pre-recorded talks with live Q&A sessions with the speakers, in a more traditional “tracked” conference

Still, let’s fully support asking questions via text chat.

  • breakout channels for workgroups/hackathon (this part can be structured like an unconference)
  • Random ad-hoc channels for open discussion that can set the current discussion topics in the channel name/description
  • Some way for people to do self-organized sessions (see the 36c3 page to get a feeling how that can look)

I think that if we are liberal enough with what we consider a presentation (and so run S-O S that is basically a talk like a talk), the rest is just a bit of scaling of the same thing, no? Self-managed break-out rooms of various size and duration.

Some of breakout discussions might be planned in advance with announced time and topic, I guess it would be nice to have some kind of Wiki for the list.

FWIW there is now a public Mumble server available at
mumble://mumble.nix-community.org.

2 Likes

I registered to https://mybuild.microsoft.com/ that happens next week so see what they are up to in terms of ideas.

PS: Thanks for all the replies so far, it’s helping me form an idea of how NixCon Online could look like.

2 Likes

Weird data point: I have recently participated in an online meetup where we managed to be annoyed that Discord break-out rooms are not ad-hoc enough … and the meetup was five participants.

(Speaking of Discord, it has some interesting failure modes that are also pretty opaque to the user)

Tried to attend a Mozilla Hubs meetup — there is a participant cap (100 members?) that was exceeded, and some people (including me in «the lobby») could not hear the sound while others could, apparently all combinations of «room/lobby»:«can hear/can’t hear» existed. I guess there are limits to platform abuse for weird experiments.

Observed an interesting catch with video talks (Zoom in Chromium). The presenter had a physical whiteboard and used a webcam. In browser it is impossible to pin the presenter, so whenever someone asked a question with voice, a black screen (no video feed) with the currently speaking person’s name was shown. Just something to take into account when preparing the configurations/procedures, I guess.