On 3 July 2014 08:37, Poul-Henning Kamp <> wrote:
> Draft -13 doesn't say what WINDOW_UPDATE(0) means ?
> Should the text read "0 to 2^31-1" instead of "1 to 2^31-1"
> or should we explicitly mention it to be reserved, undefined
> or to be ignored ?

You're right, there's no explicit description here, but I'd argue that
this text:

>The legal range for the increment to the flow control window is 1 to
> 2^31 - 1 (0x7fffffff) bytes.

means that WINDOW_UPDATE(0) is a protocol error. I'd want to tear down
the stream with GOAWAY, error code PROTOCOL_ERROR (because it's not in
the explicitly enumerated reasons for using FLOW_CONTROL_ERROR). I
think we should update the spec to either allow it (as a no-op) or
reject with error code FLOW_CONTROL_ERROR.

Received on Thursday, 3 July 2014 07:49:21 UTC