https://github.com/ipfs-shipyard/service-worker-gateway/blob/master/src/index.js
this code for IPFS was written by someone who knows how to set up service workers - review and profit
- https://gist.github.com/jimpick/2df9b96d1332e5e28bb626b8071e6677
- https://github.com/jimpick/dat-node/tree/multiwriter-staging
this repo is a half-assed hacky attempt to use service workers to transparently proxy resources referenced in the main browser dom via a dat archive (hyperdb). the idea was that static html, or at least client-side js app, could include hyperlinks to data sources that may resolve via a local static web file server, but if not, the service worker attempts to resolve them via dat distributed fs and overload the specific uris with it's fetch / cache system. In this way the source code of the static html file does not need to be modifid - in theory.
- working: duplex dat archive replication browser <--> local fs via indexdb; write in one place, see it in the other
- working: service-worker proxying of non-dat-archive image assets at certain local uris (
/collections/*&/imgproxy/*). - working: service worker dat archive replicated from browser indexdb. the sw archive can access copys of the files replicated from the main broweser window's dat archive
- not working: service worker proxying of image files from service-worker dat archive
- not sure why these requests are 404ing working - maybe hairiness re: sw cache, or a race condition
- or image data content encoding / headers are wrong
I don't know what I'm doing when it comes to service workers, low-level browser cache + req/res headers + content-encoding, and I found it hard to debug the sw (certain browsers don't make it easy to be sure the sw is updated), so overall the codebase is a big mess.
if you are going to hack on it, sw.js is the service worker source that is the most up to date. It's ugly and is lacking automated build scripts to bundle deps into the it's src (no modules in sw :(.
overall I would say that this shows that service-worker proxying of dat resources is too finicky to use. Would be better to integrate dat in browser land (as it already is) with a state management library or at least a simple js script to find and transform certain uris in the main dom.
https://hypercore-protocol-proxy-test.glitch.me/
Just seeing if hypercore-protocol-proxy will work in a web page.
MIT