Bi-directional streaming is important for things like WebSockets, especially considering that we want to keep the WebSocket client and server as efficient as possible.
Streamable bodies use the stream? -> true escape hatch which reduces the complexity of supporting bi-directional streaming clients and servers.
I have a strong desire to remove this feature as the complexity is only slightly outweighed by the benefits. As an interface, it's strictly worse IMHO. In essence, the streaming interface has distinctly different requirements on the client and server side (see #stream(input) on the client side).
As an alternative, maybe we can remove "Streamable" as a concept and instead introduce some kind of hijack! mechanism for the request and response side of the connection handling. This mechanism only applies to HTTP/1 in general.
Bi-directional streaming is important for things like WebSockets, especially considering that we want to keep the WebSocket client and server as efficient as possible.
Streamable bodies use the
stream? -> true
escape hatch which reduces the complexity of supporting bi-directional streaming clients and servers.I have a strong desire to remove this feature as the complexity is only slightly outweighed by the benefits. As an interface, it's strictly worse IMHO. In essence, the streaming interface has distinctly different requirements on the client and server side (see
#stream(input)
on the client side).As an alternative, maybe we can remove "Streamable" as a concept and instead introduce some kind of
hijack!
mechanism for the request and response side of the connection handling. This mechanism only applies to HTTP/1 in general.Types of Changes
Contribution