- From: Tomasz Pluskiewicz <tomasz@t-code.pl>
- Date: Tue, 9 Jul 2019 22:29:07 +0200
- To: Karol Szczepański <karol.szczepanski@gmail.com>
- Cc: Hydra <public-hydra@w3.org>, Angelo Veltens <angelo.veltens@online.de>, Michael Pigott <mpigott@ironhorsesoftware.com>
I think you should really look at Async API first ;). AFAIK it aims to do what you describe. It it not a tool. It is an agnostic spec for describing async communication. But I do agree with the last point. There is going to be an overlap in complex systems where both means of communication are used. A bridge of sorts would be quite useful. I’m pretty sure there is related research in the REST community. Tom On 9 July 2019 at 22:19:58, Karol Szczepański (karol.szczepanski@gmail.com) wrote: > > We should make this more egalitarian. Otherwise is it currently a bus factor of 1? > > For the time being yes. > > > Web sockets is event-based communication. A paradigm pretty different from REST. > On top of that SignalR is quite opinionated. > > It doesn't matter, SignalR or Async API Initiative, and yes - event > based communication is far from REST. > What matters is on how to propagate such a knowledge to the client, i.e.: > > "Look client, here is what you need to do to subscribe for some events. > Once subscribed, you're on your on with these events, but here is a > receipt on how to get on the bus." > > In case of SignalR subscription is a matter of sending a couple of > GETs to specified URLS with some query string params making it > possible to describe such links > (current SignalR client has these logic hardcoded), but I'm quite > curious on how such an mental exercise could end up. > > Actually, telling a client that there is another pipe with different > protocol might be interesting and be of some value. > HTTP spec defines i.e. 101 Switching Protocols status. Just loud thinking. > > Best > > Karol
Received on Tuesday, 9 July 2019 20:29:33 UTC