Re: [bitcoin-dev] BIP 158 Flexibility and Filter Size

[Jim Posen <jim.posen@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1880 bytes --]

>
> I was wondering why this multi-layer multi-block filter proposal isn't
> getting any comment,
> is it because not asking all filters is leaking information?
>

It's an interesting idea, but it adds more complexity to the client and
could be added later on if clients adopt BIP 157 and complain about
bandwidth. It also derives all bandwidth gains from address reuse. So I'm
hesitant to make the complexity tradeoff for bandwidth savings due to a
behavior that is actively discouraged.

On another note, I've been thinking that block TXO commitments could
resolve the issue we are facing now with deciding between the prev script
approach and outpoint. The whole argument for outpoints is that there are
compact-ish (<1 MiB) proofs of filter validity, which is not currently
possible if the filters included prev output data. Such proofs would be
feasible if blocks headers (well, actually coinbase txs) had a commitment
to the Merkle root of all newly created outputs in the block.

This idea has been tossed around before in the context of fraud proofs and
TXO bitfields, and seems to unlock a whole bunch of other P2P commitments.
For example, if we wanted to do P2P commitments (BIP 157-style) to the
distribution of tx fees in a block, one could use block TXO commitments to
prove correctness of fees for non-segwit txs. It also enables block
validity proofs (assuming parent blocks are valid), which are not as
powerful as invalidity/fraud proofs, but interesting nonetheless.

This would require a new getdata type BLOCK_WITH_PREVOUTS or something. I
assume for most coinbase-tx-committed proposals, we'll also need a new
getcoinbases/coinbases that requests the coinbase tx and Merkle branch for
a range of headers as well. But with these additions, we could start
serving more block-derived data to light clients under the BIP 157
at-least-one-honest-peer assumption.

[-- Attachment #2: Type: text/html, Size: 5040 bytes --]