On 3 Apr 2017, at 04:39, Russell O'Connor <roconnor@blockstream.io> wrote:

On Sun, Apr 2, 2017 at 4:13 PM, Johnson Lau via bitcoin-dev <bitcoin-dev@lists.linuxfoundation.org> wrote:
• the witness of the first input of the coinbase transaction MUST have exactly one stack item (the "extended header"), with the following data:
• bytes 0 to 3: nHeight MUST be equal to the height of this block (signed little endian)

Someone told me a while back that it would be more natural if we move the nHeight from the coinbase script to the coinbase locktime. Have you considered doing this?

Yes, it’d look much better. But I’m thinking of a different approach: instead of using a hash of 0000….0000, we use the hash of previous block for the coinbase input. With some new SIGHASH design, this allows people to pay to a child of a particular block. This is actually implemented in my spoonnet2 branch. I’ll describe it with a BIP soon

However, what I’m trying to do in the extended block header is independent to the design of coinbase tx. Here I’m trying to let people knowing the height just by a header and extended header (<300 bytes), without requiring all headers in the history.