## Summary Addresses #35. Maybe it could be wise to do some caching of the spine sizes (but performance isn't too bad).