David Faure corrected me about the KMimeType/QMimeType conversion (now stricken above), he said:
> This is a major misunderstanding. As I explained in my own presentation, the
> existing code (which I wrote)
> - didn't have a license that allowed moving the code into Qt (no CLA)
> - didn't have an API that was consistent with the rest of Qt (pointer-based
> instead of value-based)
> - didn't use the most performant technical solution (mmap'ing the binary
> cache).
>
> For all these reasons, I completely rewrote the mimetype code in order to
> submit it to Qt, with help from other people for the API.