The core idea is simple: make the resources you need to render a Lexicon (or part of it) available as blobs, and reference them from an event that states the lexicon it handles. Index those and you know what to pick dynamically. The execution is a little tricky to make safe, but I have…
…a prototype for that part. Happy to share or discuss!
@syntacrobat.xyz made a client side cloudflare workers blog that is very efficient for blobs at snowpo.st - it renders markdown to html then pushes it into the CF global cache. Should be able to follow this pattern.
Right, that's a very cool approach, but things get trickier when you need interactivity and potentially arbitrary HTML. You *do not* want to allow arbitrary embedding, you want to be super careful with JS, etc.
Yep this is “just” a recent engineering implementation for fast blobs.