Three essays
Longer arguments that did not fit in the handbook. What linked data is actually for, why a credential is a promise about a context, and why a graph needs two pictures rather than one.
Two decades of semantic web promises have produced a lot of scepticism and some genuinely load-bearing infrastructure. The useful question is not whether linked data won, but which of its bets pay off at your scale — and the answer turns on one thing: whether your identifiers have to survive leaving your database.
A verifiable credential is a signed JSON-LD document, which means the
signature covers the graph, not the bytes — and the graph is produced by
resolving a @context at some URL, later, on someone else's
machine. Everything difficult about VCs follows from that sentence.
Every JSON-LD document is simultaneously a tree a developer types and a graph
a machine reasons over. Most tools show you one. The gap between them is where
@container, @nest and scoped contexts live — which
is to say, where the bugs live.
