I think it might be way easier to configure geogebra using config.nix rather than building an overlay to pass the modified language parameter to the derivation.

You can just override the derivation. Whether you put the override in a overlay or directly in your config.nix you can decide.

The config set is for configuration of the Nixpkgs package set. Configuration of individual packages should not be in there, at least not in my opinion.

thanks for your response.

There are two reasons why I implemented this change:

• it feels (at least for me, but I can be wrong) way more convenient to adjust configuration parameters using the config.nix rather than overriding derivations to modify parameters.
• I've seen this pattern used by several other packages (e.g. aegisub, avahi, blueman or even chromium) so I assumed this is a common approach :-)

it feels (at least for me, but I can be wrong) way more convenient to adjust configuration parameters using the config.nix rather than overriding derivations to modify parameters.

I think it is in principle fine to provide functions in Nixpkgs. You could have a function there that takes the arguments you want to present and build a derivation from that. This is often not done because, while it could be more convenient from a UI point of view, it requires more effort to create and maintain such a layer, while in the end it will be hardly used. This is also why it is typically recommended to use overrideAttrs to override build flags, instead of us presenting every possible option as a function parameter.

I've seen this pattern used by several other packages

Unfortunately, yes...

Thanks for the clarification @FRidh.

As I understand the original intent of config.nix I think it's better to close this for now :-)