What are the bad things about Kord?

[HopeBaron]

Hello Team,
I'm putting this up because we are going to dive into threads and I don't want to forget about it.

We all have used Kord by far in many ways that it is powerful in many areas. However, we know that Kord is not as complete or battle-tested like D4J or JDA

I'm sharing here things that I believe could be better, please make sure to share yours as well.

Embrace open as a tool

We have lots of sealed/closed classes which makes our type system one of the very strong once in the field. However, lots of times it causes headaches to large-scale bot devs or library maintainers.
The initial reasoning behind this was to avoid instances were the user may call a function that doesn't work on this type.

While this is true, the user can make use of our existing implementations, #733 #667 are two examples where our sealed/closed fall-short.

There are many instances were we've asked our users to wrap our classes instead of inheriting from it but this comes at the cost of maintenance and flexibility. We've seen lots of users (specially from our Discord server) trying to use Kord to their hearts content, but they stop half-way through due to the hacks they had to do to get things to work.

Let's use open on classes that the user may reuse.

Cache

The cache project is used by both the dev team and the users, and it offers a variety of cache implementations that the user may use. But it falls short when it comes to complex queries or instances where the user would like to have their own custom function/ query on a certain object. #622 has been lying around for a while now, and is a good example of this case.

[luisfbl]

It would be an important point to try to decrease the number of pull request and issue, I bet kord would be better off with contributions of this kind before adding new things

[gdude2002]

Project & Development

• Kord is still making use of the legacy Sonatype Snapshots repository
• Kord as a project is very complex (though perhaps necessarily), and I find it difficult to understand or contribute to
• Kord has a development process that is, in my opinion, too formal - requiring a pile of official reviews for small changes is not going to scale and already causes problems with the project's velocity
• Kord has a low bus factor (two, both of whom have been away for some time now), and it needs to have responsibilities more spread out to combat this
• Larger, more difficult issues don't seem to get addressed
  • For example, typed message classes, generic event data extraction, and the other things we've talked about that KordEx could really do with
• It takes a very, very long time for PRs to be completed and merged, especially as larger features are prioritized over interim workarounds when new Discord features appear
  • For example, forum channels - it would have made sense to start with the minimum needed to prevent bots from breaking before implementing them properly, so that they can continue to work properly even if they can't directly make use of the new functionality

Community Spaces

• The community server is not managed competently and lacks any form of moderation team, proper community management, or moderation tooling; it would be very easy for a small group of raiders to make the server completely unusable in a matter of minutes
• For Hope specifically: Don't forget to re-read your DMs and the discussion you had with Bart about them; I still consider this a huge intersectionality problem, and I am not the only one with similar concerns

[HopeBaron]

Those are critical points, thanks for the heads up will take them into consideration

[MrPowerGamerBR]

My goal was this: I want to restart my bot quickly to be able to deploy new features and bug fixes without annoying the users, I hate restarting my bot because it consumes gateway identifies and takes too long for all shards to login to the gateway.

So I ended up attempting to migrate my bot to Kord, after all, Kord is extremely modular so it should be easy to do what I want!

Kord, while it works for small projects, it has a lot of issues when working with it on big bots. When I was attempting to migrate my bot to Kord, a lot of times I kept thinking about "well, is this my own fault, or is this Kord's fault?". Compare that to JDA, where JDA, while it has its own issues (not modular, can't use REST only, etc), it is extremely battle-tested because tons of big bots use it.

Here's some issues off the top of my head:

1. Some gateway connections just flat out die for no reason. Sadly I don't remember if the connection was stuck in a active connection or if it got stuck while attempting to reconnect. Also, the default behavior of "only attempt to reconnect up to X times" is very user unfriendly.
2. The current REST rate limit implementation is borked.
3. Kord always implemented and merged new Discord features faster than JDA. But nowadays JDA is implementing new features faster than Kord.

There were also other issues on my end too, like how it is hard to migrate a big bot like Loritta from JDA to Kord, so I ended up unknowingly adding bugs and breaking features during my migration process.

Because I forked Kord to attempt to fix these issues, I also ended up thinking about "I wonder if it is easy to fork JDA to do what I want instead", so, after a few hours, I already had a proof of concept that implemented what I wanted, and that's why I switched back from Kord to JDA.

[HopeBaron]

I'm aware of your experience this is why I asked for your feedback
Really appreciated