@pfm I actually should probably just toss the canonical s-expressions representation of linked data stuff out there somwhere, it's so damn easy to write and use and also is a way simpler canonicalization algorithm than the rdf graph canonicalization algorithm (plus it preserves the tree shape, though that may make RDF nerds angery)
@pfm "CapabilityPub" writeup coming this week.
Probably I wouldn't ship data over-the-wire with the canonical s-exps representation unless a client asked for it with the HTTP Accept header, I'd just serve json-ld. It turns out the canonical s-exps approach is great for... canonicalization though.
Octodon is a nice general purpose instance.