Date: September 9, 2026
Under EIP-1559 the base fee is a floor. Every included transaction pays at least base_fee per gas, no sender can offer to take a later position in exchange for paying less, and a transaction whose max_fee falls below the base fee is invalid before it reaches a block.
The protocol also observes willingness to pay rather than temporal preference. A liquidation that loses its value within seconds and a treasury transfer that would be equally good three blocks later can submit the same bid, and nothing distinguishes them. The second has no way to say so and no way to be paid for saying so.
This post describes one field that changes both facts, at the smallest scope I could find: one block, no new protocol state, no deferral to a later slot. It is a mechanism sketch, not an EIP. The general idea it reduces from is in RN-12, which treats temporal flexibility as something that can be supplied, demanded, verified and cleared against a neutral baseline.
What the demand evidence is, and is not
The activity mix differs across chains in a way that is hard to explain by speed alone. In the snapshot RN-14 uses, Solana’s DEX volume exceeds Ethereum’s on about a twelfth of the TVL, while Tron carries 1.43 times Solana’s active addresses on roughly one fifty-seventh of its DEX volume. That is consistent with trading concentrating on one chain and payment-like transfers on another.
RN-14 is observational. The mechanism below does not depend on it. Fee level moves flow and is the simplest competing explanation; distribution, incentives, regulatory geography and Ethereum’s rollup strategy are others.
The mechanism
RN-15 adds one signed field, a Temporal Liquidity Authorization:
TLA > 0 funding leg: authorize a payment, seek an earlier intra-block band
TLA = 0 neither leg: ordinary EIP-1559 treatment, unchanged
TLA < 0 supply leg: commit to a later band, become eligible for funding
The sign selects which leg of an exchange the transaction is on. Positive TLA is the temporal liquidity funding leg, negative TLA the temporal liquidity supply leg. They are opposite sides of a trade, not positive and negative amounts of one quantity: they are not in the same unit, there is no meaningful signed total of them, and they meet only through allocation and settlement.
The funding leg is money. A maximum lump-sum authorization in wei, not a fee per gas. Included consumers fund a block-local pool and receive earlier bands. Because it is a lump sum, a consumer’s contribution does not depend on its own realized gas, which keeps consumer gas uncertainty out of the clearing.
The supply leg is not money. What it supplies is a willingness to accept later treatment, not execution capacity. Its sign opts a below-base-fee transaction into provider treatment; its magnitude commits the transaction to a later band if included. It does not state or cap a subsidy. A larger negative number buys a later position, not a bigger discount.
Provider treatment requires both conditions:
TLA_i < 0
max_fee_i < base_fee
A transaction below the base fee without the opt-in stays ineligible. One that can pay the base fee gets no subsidy for signing a negative TLA, because it has no shortfall to close.
For provider i the builder chooses and publishes an effective tip p_i within the sender’s authorization, and the shortfall follows:
0 <= p_i <= min( max_priority_fee_i , max_fee_i )
s_i(p_i) = base_fee + p_i - max_fee_i
p_i may be the full max_priority_fee or less. A higher tip raises the provider’s shortfall and so its claim on the pool, and because the pool is consumed at the gas limit while the builder earns on realized gas, a revenue-maximizing builder does not always take the maximum. Whether that choice should stay with the builder or follow a protocol rule, proposer scoring or a cleared rate is examined in RN-15 sec. 8.
The provider pays its own max_fee per unit of realized gas. The consumer pool supplies the shortfall, completing the burn and the selected tip. The provider receives no transfer. Its benefit is conditional inclusion below the block base fee, which EIP-1559 forbids today.
That is why the discount has to reach below the base fee rather than living in the tip. For a low-value transfer during congestion the barrier is the burn, and redistributing only the priority fee leaves it standing.
Reserve, settle, and order
Realized gas is unknown during block construction. RN-15 does not estimate it. It reserves against the gas limit L_i and settles against protocol-metered gas g_i:
reserve R_i = L_i * s_i(p_i)
settle S_i = g_i * s_i(p_i)
Because g_i <= L_i, settlement cannot exceed reservation, so no estimate enters the consensus rule. Unused reservation reduces consumer charges and is refunded, and nothing is carried into another slot. An out-of-gas provider settles at its reservation and releases nothing, so it cannot draw more than was reserved for it.
Funding order and execution order are separate:
- Provider funding uses ascending shortfall
s_i(p_i), subject to the pool covering each gas-limit reservation. - Execution ordering uses
band_i = floor(TLA_i / tla_tick), larger bands earlier. Consumers occupy earlier bands, zero-TLA transactions the zero region, funded providers the later ones.
Builders keep ordering freedom inside a band, which leaves bundle adjacency workable when a bundle’s members share one.
The ordering rule is not optional. If positive TLA changed payment but not position, a consumer would pay more for the service it already had, zero authorization would dominate, the pool would empty, and no provider would be funded. Private position sales that route around the bands produce the same collapse. There is no partial-credit version of this.
What changes, and what does not
RN-15 does not change the gas limit, the gas schedule, or what a unit of gas represents. What it changes is which transactions are admissible: a provider with max_fee < base_fee is invalid today and becomes valid when consumer authorization covers its shortfall.
Whether that is a gain depends on what the provider’s gas replaces, and two scarce quantities have to be kept apart. Physical gas binds only when the block reaches the limit, and the base fee exists to hold usage at the target, which is half the limit, so a full block is the burst case rather than the normal one. The consumer pool binds whenever authorized funding falls short of what admissible providers ask, which can happen in a block that is half empty.
- Idle-capacity admission. The provider uses gas the block would have left unused. With the controller holding usage near target, this should be the ordinary case.
- Funding-rationed exclusion. The pool runs out before the admissible providers do, so some are not funded although gas remains.
- Gas displacement. Only at the limit. What is pushed out is whatever sits at the margin of the builder’s selection, which need not be an ordinary transaction: it can be a neutral transaction, or another provider that lost the shortfall ordering.
- Ordering only. The included set is unchanged and only positions move.
- Induced demand. Applications change behavior once the mechanism exists.
How often mainnet blocks actually reach the limit is unmeasured. These cases cannot be ranked by transaction count, gas or burn; RN-13 Part II sets out the benchmark: admitted and displaced demand by class, temporal and allocative welfare, real resource cost, payments and burn, builder incentives, fairness, and dynamic stability.
Several cautions
Additional burn destroys ETH, so any holder benefit runs through net issuance, is diffuse, and is shared with parties that never adopted the mechanism. And a higher fee reveals private willingness to pay, not social value.
One further question the note raises: should provider gas count in the base-fee signal the way ordinary demand does? A provider is admissible only while consumer funding covers its shortfall, so that gas is self-limiting in a way unfunded demand is not. With no pool in the next block it is not includable, and the unchanged base fee already excludes it.
RN-15 proposes one source-aware rule on that basis, treats it as a shadow-controller experiment. Validators would still execute and meter the full physical gas; only the controller’s input would change. The rule and its alternatives are in the note.
Contextual validity: a below-base-fee transaction becomes valid only with sufficient same-block funding, so includability depends on block composition.
A second authorization:max_fee still caps the EIP-1559 gas component while positive TLA separately caps the temporal charge.
Why this stops at the block boundary
The scope is deliberate. RN-15 keeps the difference from current EIP-1559 as small as a two-sided mechanism can: one field, one block, no protocol-held balance, no change to the base-fee rule or the burn.
Both legs must appear in the same block, while a consumer and a provider that would have traded one slot apart do not trade at all. Relaxing either means letting temporal liquidity persist past the block, with protocol-held state, ownership and withdrawal rules, and a second controller interacting with the base fee. RN-16, a working draft, takes that on for adjacent slots: the funding leg is money and can be carried as protocol state, while the supply leg becomes pending provider transactions in builder-local pools, which is not consensus state and need not be the same set for every builder. Further out, RN-17 asks what happens when the unit of demand is a stream rather than a transaction, with a cadence, a deadline or a bounded delay across many slots; it is a research agenda, not a proposal.
Both are mentioned only to mark the boundary. Nothing here depends on them.
Questions for review
- Can signed-TLA ordering be enforced against private position sales and bundles?
- Who should select the effective provider tip? Builder discretion, a protocol cap, proposer scoring on inclusion, or a cleared rate. A highest-bid proposer auction rewards monetizable block value, not provider count or temporal welfare.
- Should conditionally funded provider gas count in the base-fee signal the way ordinary demand does? If it should be discounted, on what rule?
- What is the right benchmark for comparing idle-capacity use, funding-rationed exclusion, gas displacement, temporal value, builder revenue, burn and fairness? And how often do mainnet blocks reach the gas limit, which decides whether displacement matters at all?
The mechanism, accounting and simulations are in RN-15; every figure here is generated by the scripts in sims/ and asserted as a test. The demand observations and their limits are in RN-14. This follows the earlier discussion of heterogeneous demand and Ethereum’s single execution lane. Notes are CC BY 4.0, simulations MIT. Any EIP would be a separate document under CC0.