I am not sure if the original decision was made to speficially exclude a pointer based system, or it just happened in the evolution of Calibre. However it is now an embedded asumption and thus hard to change.
Reasons I can think of off-hand include:
- If links get broken then you get serious problems with the base Calibre dataconsitency. It is too easy to do this if files are stored outside Calibre control.
- The databse recovery capability would not be possible if Calibre did not know where to find the metadata.opf files stored with each ebook
- Any solution has to be portable across all supported OS (Windows, MacOS, Linux)
I am sure there are many others.
All the main Calibre functionality is exposed via the Calibre CLI so it might be possible for a really interested developer to do something without having to re-invent all the capability.