Do we need tx proofs?

[RCasatta]

In https://github.com/mempoolco/drafts/blob/main/utxo-torrent/utxo-torrent.md#assumptions "we assume a set of related proofs for the UTXO is 4x."

Do we really need tx proofs?

There is no UTXO commitment, so we are trusting the UTXO provider anyway at this point I think the trust should be used also for tx existence.

Moreover, even with tx proofs a malicious UTXO provider may trick the client without producing pow (this is true also for electrum style wallet):

• The server omits previously spent transaction -> client shows a greater than real balance
• The server omits previously received transaction -> client shows a lower than real balance

[gdassori]

ACK

• Multiple authorities of consortia will sign the same work (shards of the utxo, and patches).
• Multiple clients, even the common ones not part of consortia, will share work or can produce it if in full validation mode.
• Consortia signatures will be shared across torrents.
• Actors of the consortia sharing wrong signatures will be highlighted, cause all the work at some height must hash in the same way.
• Clients will download signatures from many authorities, and work from random nodes, and can achieve consensus by downloading multiple signatures for the same piece of work from different consortia actors, even part of a different consortium.
• Just a single member of a consortium can commit his bandwidth to share the work across common nodes or to not share it at all if the network is healthy enough to take care of itself with common nodes. All the others can just share signatures.
• With a healthy network, actors of the consortia can just share signatures, certifying that the work is good.
• Cooperative nodess will share only signed shards & patches.
• Malicious ones sharing non-compliant or fake data will be emarginated.
• Any client could choose to sync with a consortium or with another. All the independent consortium will sign the same work.

Refs:

• distributed agreement on data #3: agreement on data
• patch protocol #2: patch protocol

In this scenario, seems that the proofs are useless. This brings the whole protocol to rely on a 4MB file per shard, and nothing more, so, as a recap:

• 8 shards downloaded and shared, means to observe ~ 640.000 scripts.
• it should take ~50MB to have all the 8 shards.
• with 50 extra MB a client would share 2:1 the needed data, observing ~ 1.2M scripts, improving the privacy and the network health.