Draft BIP for User-Defined Transaction Flags Policy & Strategy

[Bitcoin Error Log]

Posted here: https://github.com/BitcoinAndLightningLayerSpecs/balls/blob/main/balls-00138.md

Full text here:

BIP: XXXX
Title: User-Defined Transaction Flags Policy & Strategy
Author: John Carvalho
Type: Standards Track
Created: Apr 15, 2024
Status: Draft

Abstract

This BIP introduces a utility-optimized strategy for Bitcoin mempool policy with new transaction signaling mechanisms, including Do-Not-Replace (DNR) and Replace-by-Fee (RBF), to enhance control over transaction handling and improve the network's economic efficiency.

Motivation

Enhancing user autonomy and network efficiency through precise, user-defined transaction signals that integrate seamlessly with Bitcoin's decentralized nature and existing economic models.

Specification Transaction Signals

• Do-Not-Replace (DNR): Ensures transactions are not replaced once broadcast. This flag is encoded using a specific bit in the transaction’s version field, similar to RBF, but with inverse logic.

• Replace-by-Fee (RBF): Allows the sender to signal that the transaction may be replaced by another transaction with a higher fee. This mechanism is used to increase the likelihood of a transaction being picked up by miners in conditions of high network congestion, ensuring timely processing.

Encoding

The new flag signal, DNR, could be encoded similarly to existing RBF flags, with complementary mempool handling and conflict-resolution logic for default local enforcement.

Rationale

Addresses the need for predictable transaction handling while respecting the decentralized, incentive-driven nature of network participants.

Note: This proposal only discusses subjective, arbitrary mempool policy and handling. It is assumed that any local policy that enforces preferred hardware limits is out of scope and remains separately necessary.

Strategic Options for Mempool Evolution

There are three strategic options for evolving the Bitcoin mempool management, where only one should be optimized:

• User-defined (The ideal, optimistic option): This approach involves creating and default-obeying various transaction flags like RBF and DNR to facilitate specific goals of transactors. The primary tradeoff is that these flags are suggestions and can be overridden by miners, which means they are not enforceable but serve as strong hints to improve transaction predictability and network efficiency.

• 
  

• Node-defined (The chaotic, centralizing option): This strategy would encourage third-party mempool providers to implement their subjective preferences on transaction facilitation. The significant tradeoff here is the potential fracturing of the mempool and private, mining-pool-centric inclusion requirements, which could lead to increased centralization and censorship.

• 
  

• Miner-defined (The rational, pessimistic option): The final strategy involves removing all policies and flags, allowing miners to decide based on transaction fees or other out-of-band terms. This approach simplifies the network at the cost of significantly reducing the utility for users who may need special handling for their transactions.

Arguments for User-Definition

Option 1 is favored here because it provides a balanced approach that enhances user experience and network functionality without overly complicating the Bitcoin protocol or risking centralization. By standardizing flags that indicate user preferences, we can achieve greater harmony and utility within the Bitcoin network, supporting diverse user needs while maintaining decentralization. 

More importantly, we may be able to prevent mempool fragmentation and privatization to miners and pools for direct transaction inclusion by intentionally supporting flags that better compete and match transaction use cases within the open mempool network instead of censoring them arbitrarily.

Economic Implications

The introduction of these signals could influence transaction fee markets and network congestion patterns:

• DNR potentially improves next-block fee competition and improves network throughput by providing clearer signals about transaction permanence and relevance.

• RBF allows for dynamic fee adjustments that can enhance the certainty of transaction confirmations during peak times, benefiting users who need timely processing.

Do-Not-Replace (DNR) Use Cases

DNR is valuable in scenarios where transaction finality is crucial upon submission, without the risk of later alterations due to increased fees. Here are some specific use cases:

• Point-of-Sale Transactions:

  • Example: Retailers or service providers accepting Bitcoin in a face-to-face setting need transactions to be final immediately to prevent fraud.

  • Usage: By using the DNR flag, merchants can ensure that once a transaction is broadcast, it cannot be replaced, thereby securing the payment process at the point of sale.

• Wage Payments:

  • Example: Employers paying salaries in Bitcoin require certainty that the transaction amounts cannot be altered once sent.

  • Usage: DNR provides employers the confidence to execute payroll transactions knowing that the payments cannot be replaced or canceled, ensuring employees receive the exact intended amounts.

• Automated Payments for Services:

  • Example: Subscription services where automated payments are scheduled and should not be subject to change once initiated.

  • Usage: DNR can be applied to ensure that automated recurring payments are processed without the risk of being replaced, thus simplifying financial planning and contract enforcement.

Replace-by-Fee (RBF) Use Cases

RBF is essential for transactions where timing and confirmation speed are more critical than the immediacy of finality. Here are applicable scenarios:

• High-Frequency Trading:

  • Example: Traders on cryptocurrency exchanges who need to rapidly adjust their positions based on market conditions.

  • Usage: RBF allows traders to increase the fee on a transaction if it's not getting confirmed quickly enough, enabling them to ensure timely executions in response to market movements.

• Emergency Service Payments:

  • Example: Payments for time-sensitive services, such as premium fast delivery or emergency technical services.

  • Usage: When quick service delivery is critical, RBF enables the sender to increase the transaction fee to speed up the confirmation process, ensuring that the transaction is prioritized by miners.

• Bidding in Auctions:

  • Example: Participants in online auctions who need to ensure their payments go through before the auction closes.

  • Usage: Auction participants can use RBF to adjust their transaction fees to outpace other transactions in times of network congestion, securing their winning bids.

• Dynamic Fee Management for Wallets:

  • Example: Users sending non-urgent transactions who want to minimize fees but are willing to increase them if network conditions change.

  • Usage: RBF provides flexibility, allowing users to start with a lower fee and only increase it if the transaction confirmation is delayed, optimizing their transaction fee expenditures.

Adoption and Transition Strategy & Requirements

It is implicit, until now, that within this strategy is a requirement for Core and other implementations to abandon strategies within Option 2, by specifically removing and rejecting policy tools like mempoolfullrbf, or other attempts to overrule, filter, or otherwise filter and hamper the propagation of valid, non-destructive transactions.

This proposal is presented to the community for feedback, focusing on gathering input from wallet developers, miners, and node operators to ensure broad support and understanding of the benefits and implications of these new transaction signals.

[Peter Todd]

On Sun, Apr 14, 2024 at 08:09:51AM -0700, Bitcoin Error Log wrote:
> Adoption and Transition Strategy & Requirements
>
> It is implicit, until now, that within this strategy is a requirement for
> Core and other implementations to abandon strategies within Option 2, by
> specifically removing and rejecting policy tools like mempoolfullrbf, or
> other attempts to overrule, filter, or otherwise filter and hamper the
> propagation of valid, non-destructive transactions.

For the record, I will continue to distribute both my Libre Relay¹ and Full-RBF
Peering² forks of Bitcoin Core, and they will continue to propagate full-rbf
replacements. I'd estimate roughly 30 nodes in total on the P2P network are
running one of those two forks. Bitcoin Knots also enables full-RBF by default,
and has something like 150 nodes.

90%+ of hash rate is running full-RBF because it's common for blocks to earn
thousands of dollars worth of fees from full-RBF replacements. The idea that
miners would give that up, en-mass, just because you asked nicely for the sake
of your broken, insecure, and unused ways of transacting bitcoin is laughable.
Miners regularly modify the Bitcoin Core codebase for their own needs;
commenting out the one line of code that prevents full-RBF replacements is
trivial.

You'd do well to stop wasting your time with this nonsense and work on
Lightning or something. As for Bitcoin Core, enabling full-RBF by default is
well overdue: https://github.com/bitcoin/bitcoin/pull/28132

1) https://github.com/petertodd/bitcoin/tree/libre-relay-v27.0rc1
2) https://github.com/petertodd/bitcoin/tree/full-rbf-v27.0rc1

--
https://petertodd.org 'peter'[:-1]@petertodd.org

[Isaac Eiter]

"Example: Retailers or service providers accepting Bitcoin in a face-to-face setting need transactions to be final immediately to prevent fraud." - Transactions can't be final until included in a block. Pretending that adding a flag to the transaction makes it "final" is going to give a false sense of security for those accepting 0-conf transactions. On-chain point-of-sale payments will never be economically feasible in a bitcoinized world. And if the point-of-sale payment is large enough to justify using on-chain, then it's also worth waiting to see it in a block.

"Automated payments for services" - If you are using recurring payments, time isn't of the essence. The guarantee of a "DNR" is pointless, there's no reason to rely on a transaction flag instead of waiting for confirmation via presence in a block.

"DNR potentially improves next block fee competition" - I'm not sure this solves a real problem.

--
You received this message because you are subscribed to the Google Groups "Bitcoin Development Mailing List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bitcoindev+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bitcoindev/cc812488-9da0-4595-be3b-bcfd7ab41106n%40googlegroups.com.

[Keagan McClelland]

Gaming this out a few iterations, I'm pretty sure a widely deployed DNR policy will result in a proliferation of direct-to-miner transaction submissions and will result in less network-wide visibility of the transaction set that is staged for confirmation. At first it seems reasonable to assume that users can help block the propagation of a hypothetical DNR replacement, but the miners ultimately are unlikely to make this choice in practice. The only argument you can fall back on here is that Miners openly defying user desires will ultimately result in stagnant or negative BTC growth which is bad for their long term, but I think that argument is pretty weak in this context.

Relying on DNR type behavior in applications will definitely be insecure, but I think fighting to do it anyway has even more distortion effects that we are unlikely to want in the long run.

Keags

[Antoine Riard]

Hi John,

Few years ago, while discussing mitigations for mempool pinning at the commitment and second-stage transaction-level, we went to consider deploying new transaction-relay infrastructure directly connecting Lightning nodes and miners mempools over LN gossips [0]. This idea was disregarded, as you're introducing miners / Lightning nodes incentives misalignment (quid if your LSP who is your channel counterparty drops your gossip with a HTLC-preimage transaction within). Additionally, mempool partitioning in the miners mempools can be always tried by your counterparty with another commitment transaction.

However, one should note that since then privileged transaction-relay API for Bitcoin applications with time-sensitive traffic requirements have become more frequent with the "so-branded" transaction accelerators, where the inclusion of a boosted transaction is only relying on the reputation of mining pools. With seeing more private / staged transaction-relay API or protocols, there is indeed a concern of loosing the notion of a common blockspace market feerate that can be estimated by all full-nodes operators.

On delegating policy-choice to the transaction-issuer themselves, I did myself an experiment "transaction-issuer selected policy limits" (e.g selecting the max standard tx weight that is accepted for a transaction version) [1]. The signaling opt-in mechanism was hacked in the nSequence field, which I found a bit gross and they are issues with this approach.

Finally, I would observe that still today's mainstream Internet packets (IPv4 - RFC 794) fields have a defined meaning (e.g type of service with priorities) even if they're not often respected by ISPs due to local BGP routing policy. In practice, I think any user-defined transaction flags proposal like that relies on respecting the finality or dynamic transaction size limits shall come with a security model relying on miners economic incentives to have the flag semantics validated.

 

Best,

Antoine

[0] https://github.com/lightning/bolts/issues/783

[1] https://github.com/bitcoin/bitcoin/issues/29454