plgg-http

The runtime-neutral HTTP model, built from scratch on plgg. Pure data and pure builders for requests, responses, statuses, methods, and the HTTP failure vocabulary — no node:http, no fetch, no DOM. Its only runtime dependency is plgg.

Why it exists

plgg-server and plgg-fetch are symmetric peers — both speak the same HTTP model. Since a peer must not import a peer, the shared model is extracted below both:

plgg ── plgg-http ─┬─ plgg-server
                   └─ plgg-fetch

This is the rule the repo follows: parallel-define small clones, but extract-below large shared vocabularies. The HTTP model is large and identical on both sides, so it gets one home here.

How the model is organized

The model covers the whole wire vocabulary as pure plgg data, grouped by concern:

• method — Method (a Box union) with its asMethod/isMethod validators.
• status — HttpStatus (a validated Box<"HttpStatus", number>).
• request — HttpRequest, with Option-returning field lookups (getHeader/getQuery/getParam/…).
• response — HttpResponse + ResponseBody (SoftStr | Bytes | Stream) and the *Response builders (textResponse/jsonResponse/…).
• failure — HttpError (a Box union) with named constructors (notFound/badRequest/…) and $-matchers (notFound$()), plus httpErrorToResponse.

Everything is pure plgg data: lookups are Option, validation is Result, unions are Box, and failures are values matched by name, never by tag string — with no platform globals, so it imports safely on a server, in a browser, or under SSR. The exact types and the full builder/matcher list live in the plgg-http source.