http2: header block logical equivalence to a single frame

Question

From the http2 spec:

Each header block is processed as a discrete unit. Header blocks MUST be transmitted as a contiguous sequence of frames, with no interleaved frames of any other type or from any other stream. The last frame in a sequence of HEADERS or CONTINUATION frames has the END_HEADERS flag set. The last frame in a sequence of PUSH_PROMISE or CONTINUATION frames has the END_HEADERS flag set. This allows a header block to be logically equivalent to a single frame.

I see how "transmitted as a contiguous sequence of frames, with no interleaved frames of any other type or from any other stream" enables "a header block to be logically equivalent to a single frame", but why is logical equivalence to a single frame so important as to veer from how the rest of http2 works?

Answer 1

Is it important because the processing of a header block changes the state of the connection held by the HPACK context.

To process a sequence like:

HEADERS[stream=3,end_headers=false], HEADERS[stream=5,end_headers=true], CONTINUATION[stream=3,end_headers=true]

you would need to wait for the CONTINUATION frame to arrive before processing the HEADERS frame for stream=5.

Now, imagine a malicious client that does not send that CONTINUATION frame: it would have the ability to completely halt the processing of frames on the server and forcing the server to buffer all the arriving frames, waiting for the CONTINUATION frame - not good.

So to make things simple in the specification and the protocol, it was mandated that CONTINUATION frames cannot be interleaved, which is a special treatment with respect to all the other HTTP/2 frames that can be interleaved.