In your experience, what is the maximum effective length of a topic?

Continuing the discussion from Threaded discussion is ultimately too complex to survive on the public Internet?:

Okay, so, we have two Discourse sites for the Fedora Project.

One, Ask Fedora is for end-user troubleshooting, and by nature the topics tend to either be short and sweet because they’re general and simple, or manageable because they’re a small number of people going back and forth on one issue still topping out at ~ 30 or so. Most “top activity” topics still tend to be in the dozen-replies range.

The second, Fedora Discussion I would, eventually, like to convince everyone ought to entirely replace our development / contributor mailing list. (This is kind of a long term goal; people have … attachments. And to be fair, plenty of good reasons.)

For a lot of conversations, Discourse is great. However, we sometimes have topics like System-Wide Change proposal: Make nano the default editor which attract … quite a lot of comments. Now, that looks positively awful in the Hyperkitty archive interface, and I imagine it’s horrific in Gmail, but with many fancy email clients (as preferred by many of our community members), it’s, well, at least working as designed.

Now, I do think that we can probably get people (and moderators, and high-trust-level users) into a better habit of splitting into new topics instead of expecting threading. But, that’s definitely a change in practice, and I don’t think comes naturally to any of a) old school mailing list users, b) old school (since the 90s counts, surely) web forum users, or c) new users. Really, it’s a new way of doing things and a new habit to develop.

Meanwhile, I was looking at the Tablo TV forum, and they’ve got some crazy topics like 'Tested' Hard Drives - General Discussion - TabloTV Community — 510 replies, and basically acting like a database (badly). But they’ve got a ton of topics like Tablo Ripper - Automatically download new recordings - Third Party Apps (Rippers etc.) - TabloTV Community which has been going since July, 2015 and has 1.4k replies, or Dolby 5.1 - General Discussion - TabloTV Community (May, 2015; 375 replies), or Antop Antennas! - Off Topic - TabloTV Community (July 2018, 305 replies).

Not to pick on Tablo too much — that’s not my point, it’s just that they happen to have these real-world examples — but, these topics seem pretty unmanageable, and while the “Summarize” button reduces these topics to 100, 77, or 63 posts, the result is pretty random and not a comprehensible summary.

So, other community-interested folks, how does this work in your community? How do you deal with long topics? In your experience, what’s the reasonable maximum length? And once a post reaches that length, do you close it (using the auto close messages count setting, or manually?), or if not, how do you handle it?


In my opinion, 100 posts is about the maximum that anyone is likely to sift through. 50 posts is probably a good number to get a broad perspective of prevailing ideas.


That’s the main goal of the Summarize button – to take a very long discussion and distill it down to the 10 percent best and most relevant posts (as measured by likes, replies, quotes, read time, etc).

Now this button won’t appear until the topic has > 50 replies, and it is a “best 10 percent” so you are still looking at minimum 5 posts in a 50 reply topic, 10 posts in a 100 reply topic, 100 posts in a 1000 reply topic, and so on. But they are objectively the best posts.

And you can dynamically expand and contract the replies to those best posts like so:

This is how Discourse is a hybrid lightly threaded model.

Always a good time to remind people they can click or tap the reply icon in the editor to convert to a new linked topic:

It’s sort of the same question we’d ask Google – how many people really go to page 2, 3, 4, 5, 6, 7, 8, 9, 10+ of the Google search results when they are searching? I know for some more difficult stuff I am looking for I will definitely go to about page 4 before reformulating my query, but I doubt I’ve ever been beyond page 10 of the Google search results in my entire life.

It’s fine to have a few larger topics, but the most important thing is to encourage creating new topics rather than piling on a small number of infinitely large topics. That’s basically … chat.

In my experience threading doesn’t solve anything in particular – it creates a million-headed hydra of infinite stuff to read. The problem isn’t so much “this would be comprehensible if threaded” but “ain’t nobody gonna read that many words”. :scream:

A couple options beyond summarize come to mind. First of those is searching within the topic like so. If it’s a long topic and you only care about the word “hello”, then search for all the posts that contain the word “hello” in that topic.

Second is reading the last 100 posts, easiest to do by entering the topic at the bottom, by clicking or tapping the post count, or last reply date marker, like so…

… and scrolling upward from the bottom. You will possibly miss some earlier activity, but the idea is that the last and most recent activity is likely the most useful and relevant.


I’m not sure of “objectively the best” is tongue-in-cheek here or not. In those threads, it didn’t feel like I was getting anything like an intelligent summary, even if those were the highest scoring.

OMG WAT???!?! I’ve been digging for the :link: icon and hitting “+ New Topic”. This is great — but is it just me not finding it? Does this need a boost in discoverability?

In general, yeah, I’m with you on the futility of gigantic threaded posts — that’s definitely not a solution. Closing the thread at an arbitrary limit seems so… 1990s forum, though. I wonder if we could, instead, as the thread approaches the “real big” point, increasingly-less-gently steer people towards:

  • considering if they really have something to add
  • using :heart: or other reactions as an alternative to adding one’s own words.
  • if they really do have something to say but it’s tangential, making a linked topic — maybe even make that the default as you get close to the topic limit.

I do feel that the most-liked posts (which our “heat” algorithm favors heavily) are generally the most useful and interesting posts in a topic. If you find that not to be the case, I’d look closely at the topic and the community. Why would users be consistently :sparkling_heart: liking posts that are … bad? Drill into specific real world examples.

(Or maybe the community doesn’t use the like button much, if at all?)


Ooh, I was going to reply to this long ago but apparently did not. Resurrecting!

In the Tablo example monster threads, it looks like the :heart: is basically just never used, for whatever reason. In Fedora, I suspect the problem is that I love the reactions plugin and you will tear it from my cold dead fingers combined with “reactions probably don’t count towards the score”[1]

So may be I should really look at that and give the summary feature a second chance.

  1. which is a reasonable default given that reactions could be negative, but all of the reactions on our site are either positive or :question: (and I probably should get rid that) — they’re really just flair for different ways of saying “yay!”. Is this terrible? I don’t care!!!1 ↩︎


I’ve definitely seen a lot of communities where the like button (whatever it is called) is just hardly ever used for whatever reason. Doesn’t matter what you name it, what you call it, what it looks like… they just can’t be arsed to “do the work” of actively pressing a button on a post to indicate anything.

This is why our Discourse metrics count passive interactions such as

  • replies
  • quotes
  • mentions
  • links
  • bookmarks

Those are all significant interactions with a post even if they are not bleeding hearts :heartpulse::wink: