I get so tired every time I encounter a project on GitHub that does what I want, but is void of any real documentation nor a treasure trove of opened issues or FAQs. (Not that the old ways were perfect, but that would be a topic all on its own.) And when a project is lacking any real platform, nine out of ten times the project links to a Discord/Matrix chat for user support and resources.

I am not looking to send a chat message for help, nor looking to become part of a community, I just want to do a simple search query to look for answers.

The things that are wrong with opaque chat platforms for knowledge transfer

The main issues I see are lacking indexability, problematic community dynamics, and no ‘outsider’ access.

Now to start of with the issue of no indexability, this one is rather straightforward. No search engine will show you that help thread in 2023 describing exactly how to deal with the quirks of the OCI container such that it follows the DNS server specified by the host it attached to. Instead, we replaced this simple concept of searching online for your issue and either finding it or not, with this:

1. Search for your problem.
2. Find no results, so think about raising an issue.
3. Discover there is no issue tracker, and you should go to platform XYZ.
4. Ensure you have an account for platform XYZ, and log-in.
5. Find whatever channel/page is used for common issues/user support.
6. Use the internal search for your problem (god forbid if you’re using Matrix).
7. Maybe find it, but if you don’t send a chat message.
8. Roll a d20, anything but 20 ends you up with an annoyed response that this has been discussed 100 times before, if you’re lucky you get a link to the previous thread. If it’s a 20 you actually found an issue.

Congrats, best case scenario you already have an account and part of the help channel, but I’d say this is not an improvement. This is with the note that I’m giving whatever internal search tool that is available is actually semi-capable of finding anything.

Now onto my second point, I’m assuming this needs little introduction, horror stories about discord communities are not hard to come by. The classic antics and power tripping mods are of course not left to just these platforms, but behaviors like this are, harping back to the indexibility, not very public. On top of problematic behaviors, a thing that I noticed is that some of these communities tend to start over indexing onto whatever biases the formed community has, and in general come with a lot of drama. Something not necessarily beneficial for a project, nor something people seeking help are looking for.

And lastly, another large barrier of entry is a lot higher because you need an account, a GitHub discussion page or Discourse forum never forces you to log-in, you can freely lurk. With platforms like Discord or Matrix this is not the case, you’ll always need the platform account, with Matrix this might be less problematic due to being able to self-host your server, and there is no single controlling entity. On top of the need for an account, getting banned means you no longer have access to any resources (but of course no one ever gets banned without due reason), and an alt-account can of course work in this case.

But in general look at all the bloat and process added to something that used to be rather simple, therefor I’d argue that maybe it wasn’t better in the past, but it was easier.

Platforms that are actually suited for knowledge transfer

In the end all we want to achieve is knowledge transfer, it would make sense to do this as efficiently as possible, and for those who want allow to do some community forming. Well let there be enough platforms, some more appropriate for one than the other, for either, but at least being open for everyone to approach.

Platforms like Discourse, GitHub discussions, GitHub/GitLab issues (or any indexable issue tracker), and of course the own website of the project, immediately come to mind. Here the community forming might be a bit lacking, but Discourse has a more informal chat interface akin to most chat apps.

Although chat is again not indexable (and has no search, at least for now), it at least pushes users to commit things to a topic once deemed important. In general the close integration with their forum makes it a lot easier to disseminate this info, so it’s heading in the right direction.

Maybe ✨AI✨ will solve our issues

Getting out of the way that with ‘AI’ I mean LLM models, more specifically encoder-decoder and enconder-only models, these could solve at least the indexibility issues. Because be it out of greed from platform owners, i.e. discord, or due to crawlers of the companies building datasets, this knowledge will most likely be collected (Whether the platform wants it or not).

This indexing isn’t perfect, but neither are our current search algorithms (partially based on encoder-only models anyway). Downside is that yes, your communication interface is locked down, and correct answers aren’t guaranteed, but at least the knowledge is no longer locked in a random chat platform.

The forum is dead, long live the forum

In all the turmoil of a changing internet, ‘content’ becoming a misnomer for AI slop, and webpages becoming more ad than info (if you are lucky enough the ‘info’ isn’t SEO optimized garbage text). There are still some interesting developments in some niche communities, like the revival of personal blogs outside of software devs, with the IndieWeb community and small hold-out forums on all sorts of niche topics. It might not be too late to stop all the knowledge that is being dumped in these mostly black holes, and if nothing else, LLMs might actually be a solution to keep this knowledge easily accessible.

Thanks to the HomeLab show for inspiring this post HomeLab Show Episode 122, Why the Fundamentals are important with special guest Veronica relevant discussion at 26:30, please disregard the date, this has been in draft for a while.