[DNSB]

Originally Posted by BWinmill

Huh? If you're talking about the computer's file manager, it could be argued that Kobo does so because it suspends access to file system while connected to the computer via USB, and rebuilds the database when the connection is terminated. It works, sometimes ...

... but not always. Try adding books by FTPing into a modded Kobo device and you'll find that reboot is necessary in order to detect the new books.

I seem to remember having to write a USB plug add, delay, USB plug remove to a hardware status file to trigger the book processing without needing a reboot. I haven't done it in a while so my memory may be wrong and it may not work with the current firmware.

Originally Posted by BWinmill

Processor and screen size aren't really an issue here. The issue is that people are typically more concerned with the book title, author, and cover than the file name. I'd have to check the specification again, but I'm pretty sure that the author and title bit can be retrieved by opening the file, going to a particular offset, doing a minor amount of parsing, then closing the file. (Covers though are more complicated.) This has to be done on a book by book basis. The database approach is different: the file is opened and closed once per query, rather than once per book. The data is gathered and returned at one time, which makes filtering more efficient. Data structures are such that sorting is more efficient, as well as the original query.

The author, title, etc. are stored in the content.opf file. I've seen several epubs where the directory structure, file names, etc. have been total garbage. Sigil actually does a fairly decent job of automagically cleaning up those files. Add in manual editing, Flightcrew and epubcheck and you can get an epub that is pretty standards compliant. If nothing else, simply opening the epub in Sigil and then saving it takes care of ensuring the directory structure is good, content.opf file has the correct file name -- I've seen multiple names such as package.opf and content.xhtml and the toc.ncx file has the correct name and structure. Doing this seems to keep my Kobos as happy campers.

For the few non-English books I have, I edit the metadata to give me a transliterated/Anglicized author and title. Calibre works very well for doing this in bulk if need be.

Originally Posted by BWinmill

Which only works if you don't want any data stored about your reading, which includes the last page read. But let's ignore the counter arguments for a moment.

For the most part, I probably agree with you on a personal level. I like bypassing the cruft and getting down to business. For me, the filesystem does that well.

I just don't think that serves Kobo's business model very well, since that model probably depends upon both readers who probably lack basic tech skills and their bookstore.

I suppose you could add a mass of little .xml files for every book carrying the same information. Doing so would make retrieving data a slow and painful process in my opinion.

Given that Kobo purchased .kepubs are stored in a single directory with a hex string as the file name, it would take a fairly major rewrite to change to a directory based file oriented system. Minor correction: if I fill the internal storage on my Kobo, it will start storing .kepubs purchased from the Kobo store in a directory on the uSD card. I don't purchase enough books from there to make this an issue except when I was testing and filled the internal card just to see what would happen.

Regards,
David