Gregory's Server
FEP-ef61 update: https://codeberg.org/fediverse/fep/pulls/455
I added a couple of sentences clarifying FEP-ef61 design goals. In particular:
1. "This document describes web gateways, which use HTTP transport. However, the data model and authentication mechanism are transport-agnostic and other types of gateways could exist."
FEP-ef61 is designed to be compatible with any transport protocol, including the sneakernet. For example, it should be possible to replace web gateways with iroh nodes.
2. Location discovery using DID services. It came to my attention that some developers are trying to implement a variation of FEP-ef61 where gateways are specified in a DID document instead of an actor document. That significantly differs from existing FEP-ef61 implementations (Streams and Mitra), and has a serious practical disadvantage: it doesn't work with generative DID methods such as did:key. Support for pure key-based identities is important for several reasons:
- It is very useful for client-to-client (#p2p) communication without servers.
- Interoperability with other protocols that use public keys as identities. #Nostr is probably the most popular, but there are many more.
- It lowers the barriers to entry for client developers, who otherwise would need to deploy a did:web or something more complicated like did:webvh.
So, don't do that.
Also added a discussion section about media access control.
If media identifier only contains a digest, the gateway can't restrict access to it. This may not be a big problem because digest is very hard to guess, but an access control mechanism still might be useful. One way to implement it is to add an 'ap' identifier of a parent document to a hashlink and make it mandatory.