I'm thinking about building a client side web app environment that blocks network access and focuses on end user controlled storage. Anyone have prior examples of something similar or just general thoughts on the idea?

Love the idea. What do you/we think about IPFS? Have always been curious about a browser that "mounts" it by default. Should work in an offline-first setting?

like this? http://hood.ie/

@Chris G I was intrigued with IPFS for a while but now I think content addressing just creates as many (if not more) problems than it solves. Yes certainly offline first at least with service workers. @Mariano Guerra I was thinking the indexeddb global could be overridden to be local or cloud or auto sync, that way keeping one standard storage interface for open source apps.

have you checked https://remotestorage.io/ ?

I've built using react/redux and this: https://www.npmjs.com/package/redux-localstorage which lets you sync your redux store to local storage. It worked pretty seamlessly. Plus if you buy into redux-sagas you can use an actor model.

@Mariano Guerra Ah interesting. That is the kind of prior art I was looking for

If you use wasm for the apps, you'd get blocked network access by default. Wasm modules can only access functions that you explicitly provide from the (js-based) environment, so it's easy to restrict capabilities to what you want to allow.

Yeah I've considered the wasm only strategy but I think there's still too much friction when dealing with the dom etc.

window.XMLHttpRequest = function() {};
window.fetch = function() {};

lol 🙂

The tricky part is stopping dom nodes from triggering IO but yeah that's the general idea

What's the goal with blocking network access?

Electron is now encouraging a strong separation between processes that talk to the network and processes that talk to the file system. (For security, to avoid inadvertently loading untrusted code over the network and letting it have access to the filesystem.)

@Jared Windover To provide a guarantee that the app isn't stealing your data

I wonder if you could implement it more rigorously with a strict CSP

Actually CSP is not rigorous enough and the spec writers have even said they don't view it as a guarantee against all ways of data exfiltration.