The purist approach might work with an SQL Server, not SQLite and would not work on FAT32. Last time I tried big binary blobs it resulted in excess RAM usage, poorer performance and a giant file, that was using MS SQL Server with NTFS.
So storing the files as-is in a directory structure is what I did. The current system aids human debugging and backups. Using the current naming allows human merging and also fixing/recovery of broken systems which the Row ID wouldn't.
I've seen backups on NTFS fail because the overall path was too long or a file name had accented characters.
So I think on balance Kovid made the right decision. Or whoever started the system for the Sony ereader; my understanding was that the embryonic library was taken over by Kovid when it only supported Sony?
Not quite related, but databases are not as simple as beginners think
https://www.explainxkcd.com/wiki/ind...loits_of_a_Mom
Quote:
|
Her daughter is named Help I'm trapped in a driver's license factory.
|