Post Processing

There is a bunch of problems that could be solved with a "post-processing" stage:

1. Orphan information: when removing the last song of an album... who cleans the album?
2. Calculated properties: like duration of a playlist when a new song is added
3. For synchronization use cases, a mechanism to get "changes since X" is needed. A trigger after operations deciding if it is relevant for synch and logging it somewhere (out of tracker) could be useful.

Conclusions

We don't need to implement a post-processing stage. We have already 3 mechanisms to solve the proposed use cases:

1. Cascade on delete will solve the orphan information problem

2. The internal hook see pvanhoof's blog). Only for restricted use cases under OUR supervision

3. If the "everything from moment X" case is needed, we could provide a library to access the binary log.