I have a bit of experience with older verdions of DBase and Clipper and I find this extremely odd.

Can you give us a bit more details? I am genuinely curious: maybe Foxpro had some fancy reporting capabilities that made it more feasible but the only type of "book" I could imagine writing with such a tool is some sort of catalog.

I.e. each product has a page (maybe with part number, pictures and price, description, maybe standardized product templates with weight, color etc.) ... then each page is a record or collection of records... and then you somehow spit out a massive pdf to send to the printers?

It does seem odd, and yet, engineers still think it's a fantastic idea. This may depend on whether you think of a file system and a table both as just a form of key/value store.

For a hyper-modern example, here's a current and popular Electron-based Markdown editor, "Inkdrop":

https://www.inkdrop.app

From the FAQ, keeping in mind this is a Markdown notes editor:

Q: Can I sync my data with DropBox, GoogleDrive, etc?

A: No. You can only sync your data with a server compatible with CouchDB. Read the documentation to learn how to set up your own sync server.

And then from the docs:

Inkdrop lets you store your notes in your own database compatible with CouchDB API instead of Inkdrop's own service. CouchDB is just another open-source NoSQL database so you can deploy it on your environment for free. See CouchDB's installation guide for more informations. Using DBaaS instead of operating database by yourself is good choice. For instance, Cloudant is one of fully-managed DBaaS providers.

Also:

You can back up all your data to your local filesystem and restore it at anytime. Inkdrop stores them as JSON files continuously while you use it. ... Since the backup data is in JSON format, it is not useful in some cases. You can export all note data in Markdown format from the application menu File -> Export -> All Notes...

To be clear, I think storing markdown notes in CouchDB is fundamentally missing the point.

But I still want WinFS ( https://en.wikipedia.org/wiki/WinFS ) and JAMStack is a thing.

Think of web content in a CMS. All the assets are in a database and they get spat out through a pipeline giving a formatted end result.

I've never written a book in a DBMS(!) but it seems plausible.

I honestly doubt that "a DBMS" (especially from the Foxbase era) is really a good choice because formatting, even very basic markdown (titles, bold...) would require custom code or extensive postprocessing.

Epic! There's really something about knowing even an old tool inside out, and being able to do amazing things with it.