A working template for a Svelte / Sapper web app which uses GitHub as the back end and Firebase for authentication. HOWTO configure and run it is here.

The "NoDB" name is a bit extreme. In practice a site would maintain its own database (or use a cloud DB such as Firebase's) for indexing, processing, performance, etc. But this template shows a database isn't actually necessary.

That may sound silly / pointless -- but a model wherein a git repository is the ultimate source of truth for user data has some interesting consequences:

• An app acts more as a filter than a silo; you can point it to any public repo, and if the data there matches the expected format, it will be processed and displayed accordingly.
• Users can log in with GitHub and subsequently have their data saved to a personal repo, using their OAuth token, instead of a common one.
• Such users literally own their data, and can decide who to share it with, using GitHub's panoply of access controls. They -- and any other people / services they choose -- can also use GitHub's collaborative tools to fork / edit / pull-request / merge changes.
• Users automatically get data version control; maintain live copies of their data at all times, without having to request it; and can easily move data between competing services, or from GitHub to another repo host.

Imagine if Facebook followed this model. Then all your Facebook data, everything you'd ever posted or commented, would be in a git repo you controlled. Facebook would have a copy too, of course, for efficient processing and indexing; but you wouldn't have to rely on them to access it. Other companies could offer services letting you mine and search your own data, and provide alternative interfaces to it.

An analogy: there are lots of services like Plaid, Mint, etc., which access banking information on your behalf, via OAuth tokens or simply using your passwords. This is an example of (the secure version of) that, treating your data like your money. Here, GitHub is the data bank, and you choose which services to share data with. You can't stop them from keeping a copy - but you can stop them from owning your data, so that you can easily move to a competitor, or use this data for multiple services at the same time. You can also move from one data bank to another with ease.

I call this the "self-hosted data" model. It's pretty surprising that I can't think of any other sites/services which use it. There's Blockstack... but there's zero reason for any cryptocurrency here. (Maybe somewhere down the road.) For-profit companies are obviously opposed to anything which crumbles the moat that keeps users with them, but even open-source / nonprofit organizations don't seem to do it.