For Railgun to work, it seems a server has to have the Railgun OS/listening agent on the server or load-balancer end. On the other end, there’s Cloudflare, which effectively only receives the “differences” in dynamic content before serving it to the end user at their POP globally.
This is because Railgun doesn’t appear to be very customizable by type of request or path. A worker would be a great place to make a request to the origin and then “know” that the origin will only send “differences” from previously known structured content.
The schema for the content that’s being compared would live in KV store, instead of memcache, which is what’s required with Railgun.
How would I approach writing Railgun on a specific path, as a Cloudflare Worker, using KV store as backing storage for comparison purposes?