But where do you get the data to create the tags and final filenames from?

User types it in.

At the time I wrote the software, the CDDB sucked considerably more than it does today, and I didn't like using any of the CDDB-enabled software. Mainly because the user-interface for correcting and editing the tags (in the available software at that time) was always terrible. You couldn't ever just type in an entire album's worth of information without having to do grunt work. For instance, having to use the mouse to switch between tracks or certain fields (keyboard shortcut support was poor), or having to dance on ctrl-c and ctrl-v to get the artist/album copied to every tune.

So the whole reason I wrote Jack was so that I could tag and encode an entire group of albums, by hand, without the CDDB, very VERY quickly. Once the albums were ripped into WAV format, I could tag a dozen albums and start the overnight encoding process in about a minute's worth of typing. I made sure that each field was selected and highlighted in the proper order, and that the Enter key always went to the next logical field. And since I was the one typing it in from scratch, it would be correct every time, without typos or capitalization errors.