Following is the current buffer policy behavior in different hosts:
Request stream’s default mode:
1. Katana selfhost : __Streamed__
2. Webhost: Buffered
3. WCF Selfhost: Buffered
__Expected__: Make Katana selfhost mode also to be buffered to keep it consistent.
Comments: Discussed in triage - we don't want to change this late in the game because we are concerned that if we enabled buffering in the Katana host that we might eagerly consume the stream in scenarios such as CORS, and if there's a Soft 404, it might break the stream for any OWIN middleware that runs after Web API.
Request stream’s default mode:
1. Katana selfhost : __Streamed__
2. Webhost: Buffered
3. WCF Selfhost: Buffered
__Expected__: Make Katana selfhost mode also to be buffered to keep it consistent.
Comments: Discussed in triage - we don't want to change this late in the game because we are concerned that if we enabled buffering in the Katana host that we might eagerly consume the stream in scenarios such as CORS, and if there's a Soft 404, it might break the stream for any OWIN middleware that runs after Web API.