Async is fun, and seems to handle the traffic well. But certain things, particularly syntax highlighting, require forking off a worker process, which does not seem very elegant. Other problems include handling async errors from pre/post download hooks. It has forced awkward things like putting "http_get" methods on the request object.
It would be much simpler to just make everything synchronous and document that noembed should be setup behind a caching proxy like varnish.
I think we could re-use most of the existing code with a few minor changes.
Async is fun, and seems to handle the traffic well. But certain things, particularly syntax highlighting, require forking off a worker process, which does not seem very elegant. Other problems include handling async errors from pre/post download hooks. It has forced awkward things like putting "http_get" methods on the request object.
It would be much simpler to just make everything synchronous and document that noembed should be setup behind a caching proxy like varnish.
I think we could re-use most of the existing code with a few minor changes.
Any feedback is very welcome! /cc @clkao @audreyt