The cost of an
unfinished exchange.

A digital exchange can protect the payment and still leave somebody paying for a transaction that never finishes. The point at which a party becomes committed matters as much as the rule for returning the purchase price.

Imagine a marketplace for transferable rights to a defined computing service. A seller offers one unused entitlement. Before a buyer accepts, the marketplace checks the seller's authority, reserves the entitlement against another sale and prepares the exchange. The buyer disappears. No payment is lost and no second entitlement is created. Yet verification consumed resources, and a reserved right was unavailable to other customers.

This is a hypothetical arrangement. It exposes a question for the Digital Scarcity programme: who can cause another party to incur costs before taking on an enforceable obligation? Earlier articles examined repeated redemption and shared allowances. Here the focus is the commercial path into an exchange, including attempts that produce no sale.

Specify what completing the exchange means

For a file purchase, completion might mean that the buyer obtains specified bytes and the seller obtains the agreed payment. For a transferable service entitlement, the buyer needs effective authority to request the performance, while the seller must cease to have the competing authority being sold. The payment, the recipient's capability and the predecessor's remaining control are separate things to verify.

Dziembowski, Eckey and Faust's 2018 FairSwap paper studies exchanging a specified digital commodity for payment. Its public abstract describes a smart contract acting as an external judge when the parties disagree, and reports formal security definitions and a security proof for its construction. The authors also report an Ethereum prototype for selling large files. The supporting implementation is explicitly a proof of concept whose contracts have not been audited.

That is a useful, bounded contribution. A successful exchange of a file does not itself show that the seller has lost every retained copy, or that a service has reassigned an exclusive capability. The additional promise must appear in the specification and in the evidence used to settle the transaction. We should not quietly turn a correct-file condition into a claim about ownership, fitness for purpose or future service availability.

For the proposed marketplace, I would distinguish an offer, a temporary reservation, funded acceptance, effective reassignment and final settlement. Those stages need not be separate blockchain transactions. They are separate obligations, and a system must say which of them has occurred when it reports success.

Find the first unrecoverable expense

Lohr and colleagues' 2022 extended paper on cost fairness separates the goods exchanged from the expense of conducting the exchange. In its two-party model, Theorem V.1 establishes a specific limit: when the initiating party incurs non-negligible transaction costs and the other party can unfaithfully leave at any time, partial cost fairness cannot be guaranteed for the initiator. The result concerns those assumptions, not every commercial arrangement.

Consider the simple sequence behind the issue. One party pays to initialise an exchange. The other leaves before committing funds. Returning the first party's unsold good does not refund the expense already paid to conduct the attempt. A later deposit cannot compensate an earlier expense if the party expected to provide that deposit never does so.

The practical question is where exposure first arises. In the proposed marketplace it could be a network fee, a costly verification request, preparation of encrypted material or reservation of scarce capacity. Each needs its own record. A supplier may choose to absorb some of that work as a selling expense. What matters is that the choice remains visible rather than disappearing behind a claim that unsuccessful trades are harmless.

The service should also separate abandonment from wrongdoing. A buyer may lose connectivity, reject changed terms or encounter a defective offer. A timeout is evidence of a missed protocol step. It does not establish the person's motive. Charges and remedies need rules that remain defensible across those cases.

Put failed attempts in the denominator

Suppose, purely for illustration, that each initiated exchange costs its initiator $0.001 before the counterparty commits. Out of 1,000 attempts, 200 are abandoned at that point and 800 complete. The abandoned attempts alone cost $0.20. Allocated across the completed exchanges, that is $0.00025 per completion, in addition to the initiation and fulfilment costs of successful trades.

If 800 of the same 1,000 attempts are abandoned, their cost is $0.80 across only 200 completions: $0.004 per completed exchange. For a hypothetical $0.01 sale, those two failed-attempt overheads are respectively 2.5% and 40% of the sale price. These are chosen numbers, not current network fees, observed abandonment rates or estimates of a deployed market.

More generally, with a fixed unrecovered initiation cost c and an abandoned share a of attempts, the abandoned-attempt expense per completed exchange is a × c / (1 − a), provided a is less than one. This is an accounting identity under the stated simplifying assumptions. It is not a behavioural model or a prediction that attempts are independent. If nothing completes, that per-completion measure is undefined even though the expenses remain real.

A low charge per attempt can therefore coexist with a substantial cost per successful trade. Deliberate repetition can make matters worse, but the accounting problem exists without an attacker. Report both attempted and completed exchanges, with the point and cost of abandonment, before comparing marketplaces on price.

A reservation consumes an opportunity

Our seller's entitlement might reserve a specific machine slot rather than an indefinitely available file. Holding that slot while a prospective buyer decides can exclude another buyer. If the first exchange expires too late, the unused right may have little remaining value. Refunding the proposed purchase price does not restore the lost appointment.

This is an application to scarce service capacity, not a result measured by FairSwap or the cost-fairness paper. Its size depends on alternative demand, the duration of the hold and whether the capacity can still be sold. An empty slot does not prove a lost sale, just as a listed price does not prove the value of the opportunity forgone.

The reservation needs a release rule that the service actually enforces. On expiry, can the original seller use the entitlement again? Could a delayed acceptance still reach a different gateway? Which event wins if release and acceptance race? If the system cannot answer those questions, an attempt to reduce waiting may create two competing claims.

The programme's chosen BSV ordering mechanism can provide a common reference for authorised transitions. Its application still needs a declared confirmation policy and a tested connection to the reservation and delivery service. The timing article explains why a transaction lock condition alone is not a service-level expiry mechanism. No BSV implementation of this proposed marketplace is claimed here.

Allocate the exposure before promising to remove it

Several designs deserve comparison. A reusable exchange service may spread setup work over many trades. A provider may quote first and reserve capacity only after funded acceptance. A separately agreed reservation charge may compensate an owner for holding a perishable slot. Existing commercial credit may make compensation enforceable without a fresh deposit for every offer.

These are design options, not proven fixes. Moving funding earlier can shift the first exposure to the buyer. A deposit has to arrive, remain available and cover a defined loss under a usable claim procedure. A membership charge can spread costs while making entry less attractive to occasional customers. A rate limit may reduce abuse while excluding legitimate bursts. The comparison should follow the cost to its bearer.

My published research on governance and rule change distinguishes open admission from authority over the rules and examines commitment problems after system-specific investment. Applied here, that raises a concrete question: who can change reservation periods, acceptable evidence or compensation rules after a trader has committed? This is an interpretation of the published institutional argument, not evidence of an estimated effect in entitlement markets.

A customer should be able to retain the terms that governed an accepted exchange. A new policy can govern new offers without silently rewriting an existing obligation. If exceptional intervention is part of the arrangement, its authority, scope and review path belong in the offer. Technical verification is useful only when the parties can determine which promise is being verified.

Test the exchange that does not finish

A proposed evaluation would interrupt the workflow before and after every commitment point. It would include a buyer who never funds, a seller who never completes reassignment, an invalid delivery, a delayed acceptance after release and a legitimate participant who misses a deadline because the service is unavailable. It would test repeated attempts from the same party and attempts that the system cannot reliably link.

For each case, record the final payment position, effective capability, retained competing authority, unrecovered fees, verification work, reservation duration and remedy. Count unresolved outcomes separately from successful refunds. A refund can be correct while the appointment has passed or another cost remains unpaid.

The contribution is an analytical extension of the programme's evaluation boundary. A scarce entitlement can be specified correctly and protected against duplicate use while its market remains expensive to enter or easy to burden with abandoned work. Examining who becomes committed first makes those costs and incentives testable. For digital commerce, that changes the comparison from the cheapest completed transaction to the complete terms on which a dependable exchange can be offered.