Discourse and Slack (or other group chat apps like IRC, HipChat, Gitter, etc)

Great insights, thanks for sharing. And before I continue, I should emphasise that these opinions are my own, and do not reflect Discourse’s strategy.

I definitely do not want to see full-blown chat built into Discourse core. It can make sense as a plugin, but even then primarily as a gated sub-community, i.e. for TL3 or moderators only. (Lounge being a chat room would certainly be interesting).

It should remain a plugin because teams that are already heavily invested in a separate, richly featured chat platform will not want to move away from that, so it’s more strategic for us to make chat plug-and-play, be it via a native plugin or tight integrations with other platforms.

This is strictly an issue of cost though. If a Slack-competitor like Mattermost introduces a hosted service based on traffic instead of seats, you’d be fine.

3 Likes

Edited the title based on @downey’s comment above.

3 Likes

I completely agree that we should not aim to insert ourselves into 1. We are not Kickstarter or GitHub and have no plans to be something like that. Though there is an open question about embedded discussion which I do not think we need to grapple with.

However, I feel there are enormous amounts of parallels between us and Slack and would much prefer an integrated solution to 2 separate solutions.

I agree that chat should not go in core for a while longer, but core should facilitate chat.

Slack’s meteoric rise is a very important lesson. It is important to remember, it was launched when there were an enormous amount of existing solutions for the exact same problem.

Campfire and HipChat predated Slack by a few years, yet failed miserably in comparison. Slack entered a completely saturated market that was not making much money and created an enormously profitable product.

This was done not by using Magic :unicorn:s

  • They had the best UI, period.

  • They had the best in class, viral marketing campaign I have seen in many years. When Slack came out people were fatigued by chat solutions, at Discourse we had already left HipChat and were using Kato. In this fatigued market you had hoards of people saying … “I am bored with chat” … “Oh, but have you seen Slack, it is really cool” … people were clamouring to get on the beta.

  • They kept refining the product, correctly and in a fast enough rate. HipChat was stagnant for over a year, Campfire was stagnant for many years. The product managers for both HipChat and Campfire simply “got it wrong”. What the market wanted was a great UI and a product that evolves. We waited a year on HipChat to grant the “privilege” of editing a typo. Emojis… Oneboxes … forget about those.

  • They sold the product … This is kind of obvious, but an interesting thing is that Stack Exchange had a public chat product that was launched more than 2 years prior to Slack and was vastly superior to all the other web chat programs around at the time. 2 years later HipChat was no where near being able to catch up with it. Still today there are quite a few feature Stack Exchange chat has that are missing from Slack. Even today on Slack it is complicated to reply to an old message, for example.

  • They embraced the web, and built a unique and consistent cross platform solutions.


From the early days of Discourse we always kept chat in the back of our minds. We considered “Chat” / “Buying and Selling” / “Questions and Answers” as different flavours of discussion. We have always been a bit reticent to deal with the “Chat” problem cause “Chat” is the Crack Cocaine of web discussion :syringe:. Civilized chat is only doable in very controlled groups. Even in small groups chat tends to bring out the worst in us. It encourages short and raw communication. You don’t have time to refine an argument when you got to keep filling the white box and hitting ENTER

That said, there are just so many parallels between what we do and Chat.

  • We support a very rich security model to facilitate proper groupings of people
  • We have the message bus that gives us live updates
  • We have a very rich Markdown engine
  • We have a great system for community flagging
  • We have bookmarks, replies, quotes
  • We have a pretty good search

The only feature that is missing from core to allow for a clean chat plugin is “presence” aka… “Who is here and who is typing”

Except for that, well, we are chat. To make us behave more like chat we simply need a different UI as the 2 existing chat plugins show.

Personally, I feel quite uneasy that so much of the world is so comfortable moving all their internal communications to a walled garden where you have to pay minimum of $12 a month per user for the right to export your data. This is a huge step back from where we were with Email.

I also notice so many of our smaller trials struggle to seed a community, chat let’s you blaze through the early days, it is perfect for small groups of people engaging in unstructured conversation.

When it comes to “integrating” chat I see a few directions I would like to explore:

  1. A richer messaging system. Messages can act like chat a lot of the time, they often involve rapid and somewhat raw conversation that are unfocused. It is a balancing act, cause it could be very onerous to be forced into chat with everyone the messages you. That said, in many cases chat can be perfect.

  2. A distinct chat UI that ships as a plugin. I feel a lot of the fear of chat is having it infect a forum with unwanted features. Trying to jam chat into a site can feel very forced. A full screen Slack like chat UI could easily act as a compliment without forcing its way into the forum UI. The big advantage here is that so many of the features are shared and there is lots of value in having a single integrated notification stream. Also, another great thing that can be achieved with this is allowing people to cleanly “break out of chat”, select a few messages and turn them into a proper topic.

I am less enthused about shoutbox like UIs. One may argue that a “shoutbox” is a perfect fit for moderators, I personally think it has little advantage over moderators having another “chat” page open, especially since notifications would easily pull them into the chat UI.

With all of this type of work we have to be very careful about timing and not spreading ourselves too thin. That said, we are thinking about this stuff and how to make some progress this year.

31 Likes

Totally agree with you. :blush: Chats are potentially dangerous in several terms of civilized discussion integrated into a community system that is focused on helpful, streamlined, well-structured threads.

Maybe a Discourse community shouldn’t be conform with additional Chat functionalities. Eventually there is a different kind of (open source) solution needed.

Over the years there a so many messangers. All of them had their good times. ICQ e.g. was very popular for casual people, IRC for experienced and IT affine user groups (moderated), Facebook Chat (with real time filtering of content / censorship) is good for gated communites (like friends, colleges). And many more.

I guess the answer is in between. The users are pointing their suitable chat solution and not vice versa. You and the whole Discourse team are doing a great job. But I think, Chats with some kind of intelligence and civiled, structured conversations requiring machine learning features. I can’t imagine, this could be possible by any small scale server infrstructure of individuals.

Thanks for your post!

2 Likes

As you said, chat for private messages or group messages up to 10 participants is enough to bridge the gap.
Chat may is great to improve on-by-one relationships with my community members

Like it! It would be nice to have

2 Likes

If you limit it to 10, you’ll immediately find a customer who needs 11.

The point of Slack is that much of the communication is opt-in. If Discourse offered chat or group messages with limited numbers of participants then the feature could go untouched.

I really like this also - great idea. We already have started using messaging to “stage” discussions for community topics, and being able to do this in a more immediate way in chat would be very valuable.

4 Likes

I run a distributed team of developers. We had a Slack account. I shut it down for reasons already mentioned. It can be an unproductive distraction. It can be addictive. It doesn’t retain knowledge. It can be closed and isolating. It can bring out the worst in us.

Basically It’s like a sexy fling that lingers on too long. It starts out being fun and ends up making you feel bad about yourself in the morning.

Initially after I banned it, it felt good. Being single (asynchronous) again was great. I could get stuff done again and focus on what was important. But being single starts to wear on you after a while. You miss that human connection. You start to feel you’re missing out on something that everyone else is enjoying. And perhaps the cycle begins again…

How do you break out of that? Ideally you find someone who you have a connection with, but doesn’t drag you down. A supportive relationship that lasts.

Just like I want to believe that long term relationships can be fruitful, I want to believe it’s possible to combine chat and long form discussion. I think @sam is on the right track. Just like a good long term relationship, perhaps you need:

  1. Positive independence. You like each other, but need your own space sometimes to stay sane. The UIs shouldn’t be too intermingled.

  2. Complimentary strengths. Being good at different things can be great if handled correctly. The features shouldn’t conflict.

  3. Openness. A free flow of communication helps. Shared notifications, easy transport of content.

  4. A positive tone. Being negative all the time, even if its shared, can drag you down. The tone the product sets is important.

I think Discourse has a great chance of being ‘the one’. The one that manages to tread the precarious path to long term happiness. In addition to the reasons @sam gave, I would add that Discourse has a head start on that elusive fourth element.

14 Likes

A post was split to a new topic: Speccing out a full page chat plugin

IRC Integration?

I’d like to chime in about IRC. Internet Relay Chat has been around for about the same time as the Web, and is still very, very much used in the free software development community. In other words, IRC users are unlikely to move to a Web-based solution for their chat needs. I know I won’t: a team required me to use Slack, and the first thing I did was to plug into the IRC gateway so I can have it along with my other messaging systems, in my terminal.

I know there’s Mattermost and its IRC gateway, but I’d like to suggest another approach, far lighter than installing a Web-based messaging system that provides full Web fanciness.

Here it goes:

IRC Logs

A Category could maintain IRC logs, one topic per channel, with configurable host and channel.

This feature would automatically capture links and images or video if some are posted, and of course support Markdown if people use it (although it could also use the simpler IRC idiom with a single * for bold, etc.)

Providing IRC logs in Discourse would help make them searchable. Moreover it would allow Discourse users to use shortcut links to relevant IRC channels. These topics would be read-only.

Topic to IRC Channel

This would be the next step, to allow Discourse users to send to a specific IRC channel. Not so difficult using an IRC bot as a trampolineintermediary, but not so elegant. I’d rather not focus on this feature and encourage users to get a proper IRC client, but that could be useful for live support in some specific cases. Honestly I think it’s overkill and would not even try to implement it.

Matching IRC User with Discourse User

Given an IRC network with a NickServ, a Discourse user could indicate their IRC account in their personal profile so that the IRC topics could link to them directly, and so that Discourse would take into account their IRC participation within Discourse participation metrics. This is an interesting feature because more often than not Discourse is not the unique way a community has integrated in its workflow. (And this could be extended to other external services such as a wiki or a gitlab…)

4 Likes

Hi Sam, I saw this pop up on Twitter today… great post.

To comment on this:

When it comes to “integrating” chat I see a few directions I would like to explore:

  1. A richer messaging system. Messages can act like chat a lot of the time, they often involve rapid and somewhat raw conversation that are unfocused. It is a balancing act, cause it could be very onerous to be forced into chat with everyone the messages you. That said, in many cases chat can be perfect.

I wonder if there are a few other elements for chat:

  1. No topic. Messages between users are just the message. Even simpler.
  2. Smaller window for writing. Implies the message will be shorter (vs. paragraph style). This also allows the discussion to remain visible while replying (vs the input box and keyboard taking up the screen when replying) and for the ‘post’ button to be more prominent (when you start typing in WhatsApp, a little arrow appears next to the text you’ve typed).
  3. Fast notifications. On the app, the caching can mean a significant delay between new post and the app icon badge updating - vs. the sense of ‘instantaneous’ updates for WhatsApp/Slack.

My amateur understanding is that this cluster of UI elements supports the chat experience.

2 Likes

I’d like to see deep integration, perhaps with Zulip,
which is also instance based and fully open source.

1 Like

We already have Zulip integration:

5 Likes