Thing is for the use case I suggested workers, as currently implemented, would suck
I wrote a little proof of concept in Xojo using a console app (basically what a Xojo web app is) that spun up a helper app using this idea of the initial reply being a redirect
it works
Still means I cant hand off the initial socket though which in a way bites
But it can work just not quite as nicely as it could IF I could hand over the actual connected socket etc
and here kicks in what i always say… let a “real” webserver (e.g. Apache or nginx) do what it can do best… acting as reverse proxy and scaling requests between multiple (single-task) hosts.
I agree with you and that’s probably why Geoff is so defendant of API 2.0, because fundamentally deep down he knows he fucked up and it’s wrecking his company, but on the outside he has to present that everything is awesome.
I think it’s also incredibly clear that he has no idea on how to solve it. Nor does he want to hear anyone else’s idea. He’s trapped in a prison of his own making.
My offers of help are not just to improve the product, they’re to utilize a happy community to do some marketing for you. Something else that Xojo desperately needs. I’m pretty sure a developer would rather choose a tool that they’ve used the apps produced with, than a tool that says they’ve been around for 25 years, but example apps look incredibly outdated.