Some widely used libraries are still versioned 0.x, which under semantic versioning, means they might break the API at any time. Examples of such libraries:
- numba v0.46, 1.2 million downloads on conda-forge
- pandas v0.26, 6.7 million downloads
- scikit-learn v0.21, 2.1 million downloads
There are many popular libraries with dependencies on numba, pandas, scikit-learn, or countless other not-quite-as-popular pre-1.0 dependencies, that themselves have reached v1.0, guaranteeing certain API stability when the major version number doesn't change.
It would seem to me that it's not possible to make such a guarantee when the dependencies don't. Does that mean that under semantic versioning, strictly speaking, a library can only reach v1.0 when all its (important) dependencies have?