Open lbergnehr opened 5 years ago
I don't understand your point. Who thinks they cannot use it? Can you reframe your point? Maybe with an example?
My concern around the WebSub model is that it doesn't allow applications (whether desktop clients, browser apps, or servers) to talk across a firewall without punching a hole in the firewall.
A common misconception seems to be that the Client in FHIRcast (without the websocket addition) needs to be an end-user client, i.e. a thick client, a web application, a mobile app, etc. In most situations, however, it's not the user-facing client app, but rather that client's application server which is acting as the FHIRcast Client.
But, with the websocket addition, described in https://github.com/fhircast/docs/issues/33#issuecomment-425730218, the FHIRcast client and the user-facing client could be the same.
Hey @lbergnehr , I think we're about to make progress on a proposed Websocket approach. Lets include guidance in that documenation explaining that use of FHIRcast doesn't dictate application architecture.
In conversation with @lbergnehr and @Will Maethner, we discussed adding short explanatory text that the use of WebSub does require that the subscribing app have an accessible url and therefore likely has a client/server communication method. That method is outside the scope of the specification. The use of WebSub vs Websocket should be carefully considered based upon the subscribing app's architecture and the environment of the integration.
WebSub is easier to troubleshoot and debug WebSub is much less likely to encounter problems with load-balancer, rever proxies or other network appliances.
There's some confusion on the WebSub model and that you cannot use it since you're system's end-user clients cannot open a socket to retrieve callbacks. This is not the intended model -- the common model is that the application servers are communicating and then communicating with its end-user clients in a proprietary way. This should probably be clarified in the spec.