In a [2019 answer that the bot has just pumped](https://law.stackexchange.com/a/41728/41938) there is an excellent description of a way I have heard about how copyright derivation works with respect to software libraries:

> I don't know if software that imports/includes other software is considered "derivative", but legally it probably has to do with the ability of your software to use another module that performs similar functionality. If your software calls an API which is instantiated in multiple libraries, then it probably isn't derivative. If the API is unique enough to a specific library, then a court might rule it "derivative". For example, I have a program that uses either BSD-licensed Editline, or GPL-licensed GNU Readline. My understanding is that because the API that I use is the same for both, installing my software on a machine that uses Readline doesn't mean my software must be GPL. Please note that I could be wrong.

I read this to mean if Editline did not exist, and the only implementation of this API was provided by Readline, then the machine code created by linking the source code of "my software" with the Readline source code would be a derivative product of Readline.  This would require me to also distribute the source code of "my software" under the GPL if I was to distribute the binary.  The very existence of Editline, even though I did not use it, removes this requirement.

Is there anything we could quote to support (or refute) this statement that is more authoritative that open source advocates saying it online?  Is this based on laws as written, judgements or legal theories by respected legal scholars?