(Yes, I've reviewed the dozen other threads; none of them go quite where I'm headed.)
I'm contemplating several different options on how to include my physical books in my Calibre catalog. FWIW, I already have good ISBNs for all the pbooks, along with some tags associated with each.
Some of the debates I've been having, that I'd be interested in experience, suggestions, etc:
Preserving the true ISBN is important to me, for the pbooks. For the ebooks, not so much. As discussed at some length by kiwidude
here (updated link), Calibre does some aggregation/collation via Worldcat when updating metadata, and the ISBN in the ID can get updated (along, of course, with adding an ASIN that may or may not be the exact edition, etc.) I have experimented with adding my own identifier paper_isbn, and with adding my own custom column; I can argue both ways, though I'm leaning toward custom column.
More generally, is there a way to constrain the metadata updaters to be edition-specific? Ideally, I could set this based on a criteria, next best would be setting the settings and then updating just the pbooks. (For example, for ebooks, I just want a representative cover. For pbooks, I'd rather have no cover at all, if the exact edition cover is not available.)
I've gone back and forth twice on whether the same ebook and pbook should be one record or two. (And this is a sub-question of whether pbooks should be in their own Calibre library.) At the moment, I'm leaning back toward separate records in a single library.
(For example, I have two physical copies of Heinlein's Glory Road, which are distinct physical editions: 0425048659 and 0671877046. I also have an ebook, currently cataloged as 9780765312228. In my current thinking, these would be three rows, so that the two pbooks could have their own #pbookISBN column entry, and so that each could have its distinct _location_* tag.)
This does, however, make a hash of duplicate book detection, although I'm thinking that Virtual Libraries will help here. It also makes rating and marking as read (which I currently handle with a custom date column) a bit untidy. Very interested in thoughts on this.
Thanks.