Fix the Filters or Transactional Censorship?
The current debate surrounding Bitcoin’s mempool filtering policies has exposed deep divisions within the Bitcoin development community, particularly over the treatment of non-monetary transactions such as Ordinals, Inscriptions, and privacy-enhancing CoinJoin transactions. Ordinals, Inscriptions, Stamps, and CoinJoins are all valid Bitcoin transactions under current consensus rules, paying market-rate fees for blockspace, and attempts to filter them based on content or intent constitute a form of censorship driven by subjective disagreement rather than protocol-level invalidity. At the heart of the controversy is a proposal, initially introduced by Peter Todd, to remove legacy restrictions on arbitrary data storage via the OP_RETURN field. Supporters of this change argue that these filters are largely ineffective due to workarounds like direct miner submissions and alternative relay networks, and that retaining such limits creates an illusion of control without practical benefit. They contend that removing these constraints not only reflects Bitcoin’s permissionless nature but also reduces centralisation pressures caused by reliance on non-standard mempool implementations.
On the opposing side, critics argue that lifting these constraints would accelerate the transformation of Bitcoin into a general-purpose data ledger rather than a monetary network, effectively degrading its core value proposition. Concerns have been raised about increasing blockspace congestion, larger UTXO sets, and the long-term impact on decentralisation. Developers like Jason Hughes of Ocean Mining have gone so far as to warn that such changes could turn Bitcoin into a “ worthless altcoin.” More broadly, dissenters view Bitcoin Core’s unwillingness to incorporate community feedback as a sign of creeping centralisation, with some node operators shifting to alternative implementations like Luke Dash Jr’s Bitcoin Knots. This growing exodus reflects not just policy disagreements, but a deeper erosion of trust in the stewardship of the network by Bitcoin Core.
A key aspect of the conflict lies in the distinction between consensus rules and relay policies. The proposed changes do not alter Bitcoin’s consensus mechanism but modify the behavior of nodes in the transaction relay layer, specifically which types of transactions are propagated. Proponents of stronger filtering (including Luke Dash Jr., who capped OP_RETURN data to 42 bytes in Bitcoin Knots) argue that so-called “spam” and privacy-preserving CoinJoin transactions can degrade the user experience and expose the network to legal scrutiny. Others, including privacy advocates, argue that such filters amount to censorship of valid, consensus-approved transactions, creating a slippery slope where certain use cases, especially privacy-enhancing ones, are unfairly penalised or excluded from block inclusion.
The implications of this debate extend beyond the technical question of how Bitcoin nodes handle unconfirmed transactions. They touch on the philosophical foundations of Bitcoin: whether it should serve strictly as a monetary network or remain open-ended and resistant to gatekeeping. While Core developers stress pragmatism and network efficiency, critics worry that mempool policy is becoming a backdoor for soft governance. The rise of alternative implementations like Knots (and to a lesser extent, Libbitcoin), the calls for decentralising policy decisions, and the growing preference among some miners for downstream forks suggest that Bitcoin is entering a new phase, where protocol design, ideological commitments, and governance structures must be actively reconciled to preserve its decentralised and permissionless philosophy.
Many Bitcoiners Note the Similarities to the Fork Wars
The current OP_RETURN and mempool filtering debate bears many striking similarities to the Bitcoin Blocksize wars of 2015–2017, which marked one of the most divisive chapters in Bitcoin’s history. At the heart of both disputes is a fundamental disagreement over the purpose of Bitcoin, whether it should remain minimal and censorship-resistant, or evolve to accommodate broader and more flexible use cases. Just as the blocksize wars pitted those advocating for larger blocks and on-chain scaling (to accommodate more transactions) against those prioritising decentralisation and node accessibility, the OP_RETURN controversy centers on whether Bitcoin’s blockspace should be selectively filtered to exclude non-monetary data like inscriptions and CoinJoins, despite these being valid under consensus rules.
The blocksize debate began when some developers and businesses pushed to increase Bitcoin’s 1MB block size limit to address rising transaction fees and congestion. Their argument was that scaling Bitcoin on-chain was essential to make it a competitive, global payment system. Opponents argued that increasing the block size would centralise the network by making it more difficult for average users to run nodes, undermining Bitcoin’s decentralisation and censorship-resistance. This disagreement culminated in a highly contentious period, where multiple proposals, such as Bitcoin XT, Classic, and Unlimited, attempted to hard fork Bitcoin. However, these attempts failed to gain sufficient consensus.
The resolution came in the form of Segregated Witness (SegWit), a soft fork activated via the User Activated Soft Fork (UASF) mechanism. The UASF was a grassroots movement where users and miners signaled support for SegWit activation independent of Core developers or major mining pools. This approach ultimately succeeded, forcing SegWit activation in August 2017 and demonstrating the power of user-driven governance. However, the fallout was swift: those who still opposed SegWit and favored large blocks forked off to create Bitcoin Cash (BCH), marking a permanent chain split. Later, the SegWit2x compromise, an agreement to activate SegWit and then double the block size, collapsed due to lack of consensus, further solidifying the divide.
Today, the OP_RETURN filtering issue echoes many of the same themes. Advocates for filters argue they preserve network efficiency and limit spam, while critics see this as a paternalistic overreach that undermines Bitcoin’s neutrality and censorship-resistance. Just as the blocksize war clarified who controls Bitcoin’s rules, the users, not the developers or miners, this new debate is prompting many to revisit what it means to run a node and what obligations, if any, software implementations have to neutrality. With users already migrating to alternative clients like Bitcoin Knots, the parallels to 2017’s UASF and chain split moment are impossible to ignore, and the outcome may again redefine the balance of power in Bitcoin’s governance model.
How Practical is Filtering Transactions Perceived as “Spam”?
The Bitcoin mempool, short for “memory pool”, is a decentralised staging area where valid but unconfirmed transactions are held by full nodes until miners include them in blocks. Each node maintains its own mempool, meaning there is no single authoritative pool, but rather a network of independently configured ones that relay transactions to each other. When Segregated Witness (SegWit) was introduced via soft fork in 2017, it moved certain parts of transaction data, specifically signatures (witness data), outside the traditional transaction structure, enabling more efficient block space usage. Data stored in SegWit’s witness field is heavily discounted in Bitcoin’s block weight calculation, counting only one-fourth as much as standard transaction data, which inadvertently incentivises use by protocols like Ordinals and Inscriptions to embed arbitrary content at lower cost. Crucially, witness data does not contribute to the UTXO set and is not validated by non-upgraded nodes, making it a target for novel uses of Bitcoin’s blockspace, both legitimate and controversial.
Ordinals and Inscriptions, introduced in early 2023, exploit SegWit’s witness space to inscribe arbitrary data, such as images, text, or entire files, directly into the blockchain. By using Taproot outputs and the malleable, discount-weighted witness structure, these protocols bypass traditional limits and embed metadata into transactions without violating consensus rules. CoinJoin implementations, like Whirlpool, similarly utilize advanced scripting and witness space to enable collaborative, privacy-enhancing transaction coordination. All these techniques operate within Bitcoin’s rules and pay the required fees, but they diverge from conventional monetary use cases, raising concerns among some developers and miners about the long-term sustainability and “intent” of Bitcoin’s blockspace.
Ocean, a Bitcoin mining pool spearheaded by Luke Dash Jr. and backed by Jack Dorsey, implemented filters within its node infrastructure, based on Bitcoin Knots, that reject certain classes of transactions deemed “spam.” These include Ordinal inscriptions, but also, CoinJoin transactions such as those produced by Samourai Wallet’s Whirlpool or similar CoinJoin implementations based on Wasabi wallet, and Joinmarket. The filter works by imposing tight limits on OP_RETURN size and rejecting transactions that use more witness space than considered acceptable by Ocean’s policy. While Ocean argues this preserves network efficiency and upholds Bitcoin’s monetary focus, critics view it as a form of censorship, warning that it fragments the mempool, weakens network neutrality, and undermines Bitcoin’s permissionless nature. Those who oppose transaction filtering claim that filtering so-called “spam” transactions at the node or mining pool level is largely impractical, as users can easily bypass these filters by submitting transactions directly to miners or encoding data in non-filtered parts of valid transactions, such as witness fields. Attempts to enforce such filters risk fragmenting the mempool, undermining neutrality, and setting precedents for subjective censorship of otherwise valid, fee-paying transactions.
The proposal to remove OP_RETURN relay size limits would have far-reaching implications for how Bitcoin nodes propagate and store data. Currently, Bitcoin Core enforces a default 80-byte limit on OP_RETURN outputs, restricting the type and volume of data users can embed via that opcode. Removing this limit would not change Bitcoin’s consensus rules but would alter how nodes relay transactions across the network, allowing for more expansive use of OP_RETURN. This could reduce the incentive for users to hide data in witness fields, thereby potentially decluttering them, but it would also make it more difficult for node operators to filter transactions based on content, challenging current practices like those employed by Ocean. The debate hinges on whether such filtering preserves Bitcoin’s functionality or compromises its foundational principle of neutrality.
The post appeared first on Bitfinex blog.
