Running server in Proxy Mode - useful for desktops

[joepio]

The Atomic-server desktop distribution is actually often not that useful to run as a real server, since you're likely to turn off your desktop:

• Goes offline when desktop is offline, so sharing data with others is not a great option
• It's often hard to set up port forwarding / ddns stuff

However, it can still be useful to run it locally, mostly as a proxy:

• Best possible performance (0ms responses on my laptop)
• Back-up when other things are offline. True data ownership!
• Host the GUI on a desktop (better UX than in a browser)

This does, however, open up some questions:

• If it functions as a proxy, it needs to sign HTTP GET requests meant for other servers. The /path endpoint does not support non-public resources, at this moment.
• How does the client get updates about which Commits are new? How does the server construct this collection? How is authorization performed? see https://github.com/ontola/atomic-data-docs/issues/28

[joepio]

Related: #998

Syncing Atomic-server with another instance would work better if we use identifiers that are local-first, or at least scoped to a drive.

Idea

• Atomic-Desktop asks for one or more drives to sync with
• User provides their secret / authentication
• The entire state of that Drive (including all its children and files) are replicated
• Perhaps this is done through content addressing (iroh / ipfs #66 ), like having one shared iroh document.