@mro That's a fair point about design by committee, and I don't disagree that RFC 9421 added complexity without proportionate benefit. But I think it's a somewhat separate question from what I was getting at.

Even if every ActivityPub spec had been designed with ruthless simplicity—no committee sprawl, no redundant signature schemes—the DX problem would still be there. You'd still need to know what to do when a property arrives as a string instead of an array. You'd still need to handle embedded objects vs. URIs. The gap between “I understand ActivityPub conceptually” and “I can ship something that actually federates correctly” would still be wide.

My point isn't really about why the stack is complex. It's that the complexity, whatever its origin, currently falls entirely on the application developer. A framework should be able to absorb most of that—the same way web frameworks absorbed HTTP parsing and content negotiation—so developers can focus on what their app actually does.

(And yes, the CGI analogy was never meant to be architectural criticism! Just reaching for a feeling most of us remember.)