feedbackbg

In @/feedback I introduced a new commenting and reacting system to this blog. Why?

For two I'm thinking of having an upcoming experimental project where I would need some interaction from a few people over a longer period of time. But for that I need to know if anybody is consistently reading this blog and whether they would interact with the widgets here. I need to lower the activation threshold for engagement to facilitate this. The old commenting system was a high-time-investment interaction because of the @/cooldown system. I wanted to retain the cooldown system because I still think it's a reasonable idea. So I needed something else: reactions fit the bill! They are everywhere so might as well be here. I think they can be done in an useful way, and I was thinking about them for a while, see @/reactions. I guess it's time to try out the idea.

For three seeing interaction on this site makes it more intrinsically rewarding to keep posting stuff. So let me allow 3-click reactions. Will seeing such reactions motivate me to post more? And what are the reactions going to tell me?

For four I was also yearning for some account management so that I can have more generous limits for trusted people. So I implemented that too at @/account. For starters registered users have a reduced @/cooldown timeout for posting comments.

But for now I retain the possibility for the no-registration needed, anonymous commenting and reacting. If I ever reach a point where there are multiple regulars here or the anon feedback is overwhelming then I will disable that as I won't be needing it anymore.

The old @/comments system was singly threaded. It allowed me to reply to individual comments and the reply would appear next to the replied-to comment. I stole this idea from https://sive.rs where the blog's author also often provides short responses to some of the comments. But this didn't work well because once you start having a discussion, things become hard to follow.

The threads model doesn't have this problem. Each user can start a thread and other users can add replies to it. Github discussions work like this. But the actual inspiration for me was a https://gerrithub.io like code commenting system. In a code commenting system you can start a thread on any line of code and then have a context specific discussion there. I liked this type of discussion, so thought might as well be used on a regular forum too.

The other end of the spectrum is a tree based model like reddit. Each response can have a reply and the discussion is presented as a messy tree. That is very hard for me to follow and scroll through, I don't like that, didn't want that here.

The inspiration for reactions is mostly github and the fact that most discussion software have reaction support. Initially I was skeptical but then I convinced myself that it can be useful, see @/reactions. Especially after seeing github discussions where a user posts 3 comments each describing a design alternative and then the users can upvote the one they like the most. I really like how easily people can get feedback about ideas with the very little spam. Check the first 4 threads on https://github.com/golang/go/discussions/71460 for an example.

There is a lot of existing systems that I could implement comments with. Take https://giscus.app/ for example. It allows using github.com discussions as a commenting system on blogs like this.

There's also dozen systems for using a combination of bluesky, mastodon, facebook, etc. as comments. Then there are ones with a custom backend database too.

But none of these systems have the exact properties I want: e.g. the very specific reactions. And using those systems would mean I won't own the data. My moderation capabilities might be limited. And the readers would need to log in to these other systems which can then track them over websites. Or these systems might decide to lock down 3rd party integrations and with it taking away my content without me having a say in it.

In the end I'm experimenting here and having everything my own code gives me more control and flexibility.

Most sites implement reactions in an addictive way. You post something online and then you constantly keep checking back for the upvotes. I think that's a bad, inhumane design. That's how a drug dealer designs software.

The solution is simple: update the reaction counters only twice a day, at UTC noon and midnight. Checking back multiple times within a period is pointless. So this should reduce fueling addictions. And to make things more complicated and unpredictable: when an update happens, reactions from the last 6 hours don't count yet. So at 12:00 only the reactions up to 06:00 are counted.

It makes the implementation simpler too: my server needs to compute each post's HTML page only once every 12 hours and it can serve a cached version rest of the time. It's still JS that renders the widget but the counters are hardcoded into the HTML. It is JS that renders it so that it can highlight the user's own reaction. And the user's reaction is fetched as a separate HTTP fetch. But this doesn't happen for unregistered users which most traffic is anyway.

The reader needs 3 clicks to react: click the smiley (express the desire to give me feedback), click the reaction (think what feedback they want to give me), click submit (confirm the feedback). Other platforms allow like/dislike/upvote/downvote with a single click.

I decided against the single-click design in order to improve the quality of this feedback a bit. If it's too easy then it's easy to misclick without realization. And the simplified interface doesn't really make the users think about the feedback. Here you need to think hard what feedback to give. In exchange it will be more useful.

I understand this reduces the potential engagement on this site. But that's fine: I'd rather have no engagement than dealing with low quality results.

I was thinking adding a "Motivate me by giving feedback:" text before the smiley button to draw readers to its attention. But then I decided against it. I don't think I want to advertise it too much. That would be too cringy. And again, I only need consistent feedback from a few regular guests.

Maybe nobody will react at all. Well, that is good feedback too. Then I know that either nobody reads this site, or it's not worth reacting about, or the reaction system doesn't work.

I'm well aware that if I put such a widget on my blog then it might affect my voice to the worse. I might observe that some posts do better than others. Usually it's the easier to digest, reactionary, anger making posts that gets high engagement on other platforms. So I would be prone posting more such content if I'm craving some attention rather than posting more constructive words.

I'm not sure how to avoid such temptations if the reaction system is a given. One must be mindful about, hence I'm writing this to myself as a reminder. https://blog.codinghorror.com/the-field-of-dreams-strategy/ explains that some people manage to not care about readers too much. I think I'm doing good on that front, I hope this continues.

Anyway, let's see how this goes. Oh, and how I implemented this? The short version is that I'm using the free tier of Cloudflare's D1 as the backend. But I'm doing weird stuff so this is a topic for another day.

#c1 by aeioypvn-guest on 2025-05-14

From time to time I have a look at your blog. I like the minimalist design.