祝融说。

Part Three does not declare the constructor the winner; it requires the defensive scheme to accept the very standard of evidence its opponents have just used. Chapter 18 distinguished external action from internal choice and proposed that the assignment of responsibility must share verifiable material. We now turn to that layer of common starting point: how minimal facts enter contention, and how they retain the conditions for correction. After the blockade eased, the conflict in Qingya Valley shifted from external alliances to internal accounts: at the council, the committee faction and the questioning faction each cited their own archives, the same contract carried two sums, and the same rainy season had two sets of water levels. Either side could declare the other's records forged, and every such declaration made the next common discussion more expensive. A council member proposed establishing a "shared base": a list of entries both sides acknowledge, serving as the common starting point of all contention. The proposal itself was unproblematic; the problem lies in every subsequent link — what to include, who keeps it, how long entries remain valid, what to do with those who refuse it — any of which can turn the base from a common starting point into one side's position. This chapter treats three interlocking questions: by what the base's admission and minimality are sustained; how it is used as a threshold for excluding dissent, and how it is captured under a forged appearance of neutrality; and what price contention pays for abandoning the base, a price that no gain elsewhere can compensate.

The Minimal Scope of Shared Material

Even the fiercest accusation — "you forged the water-level records" — presupposes that water-level records exist, that there were days and places of measurement, and that there is a text that could be forged. The two sides of a dispute share far more at the level of objects than at the level of conclusions. A position that denies all shared objects outright cannot be argued with; it can only be ignored or suppressed, for it leaves the other side no position from which to rebut. Such a position thereby automatically renounces persuasion and retains nothing but confrontation.

What the shared base includes are claims, chains of provenance, and conditions of measurement, not the "true meaning" of events. Hourly water-level readings can enter the base; "the dam is therefore safe" or "the committee is therefore derelict" must remain at the level of interpretation. Chapter 2 already showed that the stability of public facts comes from clarity of objects, traceability of provenance, and availability of correction, not from agreement in conclusions. The closer an entry stands to the object, the more likely it is to be used in common by different positions; the closer it stands to the conclusion, the faster it becomes one faction's banner, discarded wholesale by the other. The base is therefore inherently incomplete: it does not answer why contention arises, nor promise how it ends; it only guarantees that when contention arises, both sides stand on the same ground where each can point the other out. Acknowledging this incompleteness is the first line of defense protecting the base from being required to bear the entire burden of explanation.

If every record can be voided wholesale according to position, contention no longer exchanges reasons but only volume, headcount, and position: whoever controls the council hall, whoever can mobilize attendance, whoever can hold out to the end, wins. This is not rhetorical exaggeration but a structural account of the cost of abandoning the base: when argument is replaced by force, the first to lose the standing to speak is always the side with the fewest resources, because organizing records, maintaining archives, and appearing again and again all cost money and time. Two sides that proclaim there are no common facts at all are in fact handing contention over in advance to whichever side has more resources.

The Admission Threshold for Minimal Facts

Entries in the base can first receive an explicit status, with the question of adoption taken up afterward. The table below sketches the fictional recording rules of Qingya Valley; it is not a matter of determining truth by majority vote.

Status can change with the material. Agreement of the two sides is not a sufficient condition for a fact to hold automatically, and a single source is not proof that the material must be false.

The provenance relations distinguished in Chapter 2 apply here: a record obtained separately by two mutually independent factions whose results coincide may enter the base; if only one side holds it, it must be marked as single-source. What matters is independence, not quantity: the same document relayed ten times is still a single source. Base entries should carry verifiable points of access: where the original is held, who may consult it, and on what conditions it may be copied. An entry the other side cannot actually inspect, however important, can only be registered under the status of "one side's claim," never under the status of "confirmed fact." This is not disparagement but an honest labeling of the support structure: a one-side claim can equally be cited, contested, elevated or overturned by new material; it differs from a confirmed entry only in provenance relation, not in political treatment. To conflate a difference of standing with a wholesale abolition of the difference in credibility, and to write single-source material directly as fact, are two faces of the same error.

An entry states "on March 9 the two stations recorded water levels of such-and-such, with known error"; it does not state "the water level is safe." The latter folds a measurement into a value judgment, making anyone who questions the measurement appear to question safety, or the reverse. The discipline of claim format is the base's breakwater: it directs revision toward specific links — a particular reading, a particular chain of provenance, a particular transcription — rather than toward wholesale acceptance or rejection.

The base includes not only confirmed items but also known gaps: originals lost during the emergency relocation, the two days on which the stations' records contradict each other, requisition compensation that no one has claimed. Chapter 7 showed that honest unknowns have boundaries: they allow neither side to fill the gaps with imagination, nor the other side to declare, on grounds of "unconfirmed," that something did not happen. Gap entries make "we do not know" common property rather than one side's cover or weapon.

Minimality Requires Active Maintenance

The operational criterion of minimality is this: remove this entry, and do both factions' arguments lose their footing? The contract total, the location of the requisitioned parcels, and the effective date of the emergency ordinance satisfy this condition; "the committee responded appropriately" or "the residents suffered most" do not. The former is a floor both sides must stand on; the latter is one side's position. The larger the floor is laid, the fewer the people who can stand on it; every expansion of the base pushes some portion of the people toward the parallel archives outside the door.

Either side may propose additions or deletions, but motions to add interpretive content should be procedurally shelved rather than debated and voted on: if the base could absorb interpretation by majority decision, it would become a third camp, larger than the two existing factions. Shelving is not destruction: shelved motions are recorded together with their reasons, becoming part of the contention at the level of interpretation and remaining available for citation later. Handling matters this way loses a little efficiency and gains rules of inclusion that do not compete in the factional contest.

Material that does not enter the base should not vanish silently. Rejection records — the content of the motion, the threshold clauses invoked, the decider, the date — constitute the base's boundary documentation, making "what did not enter" as auditable as "what entered." With a base that bears no trace of rejection, outsiders can see only its list of contents and cannot judge whether it truly lacks qualifying material or whether qualifying material was blocked at the door; the boundary documentation is precisely the only evidence that distinguishes these two states.

Whether compensation is adequate, whether risk is acceptable, and how loyalty should be measured are questions of value. The base can supply the material they require; it cannot supply the answer. As soon as someone begins with "the base shows" and derives a normative conclusion, listeners should examine the hidden premise: the base's authority comes from provenance relations, and provenance relations can support only factual claims; they cannot complete, for anyone, the step from "is" to "ought."

Custodianship and the Qualification to Enter

The traceable chain of Chapter 14 becomes concrete here: the preservation of originals, the authorization of access, and the adjudication of versions in the base should be held by different roles. The keeper cannot alone decide who may look; the authorizer cannot rewrite versions; the adjudicator of versions does not keep the originals. Should any role hold two of these functions at once, the chain contracts into gatekeeping power, and the intermediary failure described in Chapter 14 replays inside the base — except that this time the failing intermediary speaks in the name of "common facts."

The custodian's duty is to keep entries usable, inspectable, and correctable, not to announce what an entry means. If someone at the council asks whom this record shows to be responsible, the custodian's correct answer is to point to the material and the dispute records rather than to deliver a ruling. Once the custodian habitually rules, each faction's resources flow into contesting the custodial post rather than verifying material — precisely the substitution the base exists to prevent.

The base does not require the destruction or merger of the factions' own archives. The parallel archives are the base's sources of correction: requisition photographs kept by the questioning faction may later be added to the base, and minutes kept by the committee may refute an exaggerated accusation. Any proposal for a "unified record" should be recognized as a move of contraction: what it eliminates is precisely the independence on which the base depends, and once independent sources disappear, "multi-party confirmation" degenerates into many photocopies of the same material.

The Exclusion Mechanism of the Base as a Ticket of Admission

The most direct abuse is to make "acceptance of the base" a precondition for speaking: those who do not accept every entry are barred from the council, their testimony automatically downgraded, their material refused a hearing. The base thereby turns from common starting point into a loyalty test; the qualification-by-declaration mechanism described in Chapter 18 reappears on new ground. The beneficiaries are the faction that controls inclusion; those who bear the cost are the people holding partial dissent — especially those whose evidence happens to fall where the base does not reach, who have the best reason to reserve judgment on individual entries and are therefore the first to be expelled. The conditions for rebuilding common facts after this kind of exclusion are correspondingly strict: the expelled should be able to re-enter on the strength of specific material; their past dissent must not accumulate as qualification stains cited against them; and the periods of silenced speech caused by exclusion should be visible in the record, for a silence that cannot be quantified is the easiest thing to deny afterward.

Dissent from the base falls into at least three classes, which should be handled differently. Those who contest a specific entry and attach new material trigger review and provisional annotation of that entry, without thereby forfeiting their standing to speak elsewhere; those who deny the base wholesale yet still take part in contention should have their denial recorded, with each party weighing for itself the weight of their subsequent claims; and those who use denial as a tactic while citing the entries of the base that favor them can be called out for this selectivity, but this remains a burden at the level of persuasion, not grounds for sanction. The danger of conflating the three classes is that vigilance aimed at the third lands, in the form of sanction, on the first — and the first are precisely the base's most important source of correction.

The side asserting that someone "does not accept the base" bears the obligation to name the specific entries rejected and the form of the rejection. A general impression of the kind "he has reservations about the base" cannot serve as the basis for a qualification ruling. If a ruling is nonetheless required, then the ruler, the grounds, and the path of appeal should all be placed on record, so that the ruling itself becomes an object of review: an exclusion mechanism that accepts no review has completed its slide from threshold to purge, and a purge needs no new reason, only time.

A side that directly benefits from the base's record of inclusion — its contract recorded as valid, its conduct recorded as compliant — should not simultaneously serve as the final adjudicator of rejection disputes. This presumes no dishonesty; it removes a structural convenience obtainable without any cheating: a side that is at once athlete and scorekeeper will see the distribution of outcomes drift toward it even when every step complies with the rules. Qualification disputes touching that side's interests should be ruled on initially by roles farther from the interest, with the dissenting party's objection preserved as an annotation.

Capture and the Monopoly on Interpretation

A base is not neutral because it calls itself neutral. Its neutrality exists in inspectable relations: both factions can submit motions; rules of inclusion are published in advance and applied equally; reasons for rejection are put in writing; time limits for acceptance are open to inspection. Capture usually begins not by overturning the rules but by selective enforcement: the same rule is applied swiftly to one side's motions while the other side's motions pile up for months; the text of the rule changes by not a single word, and the base has already changed hands.

What counts as an "event" for inclusion already carries interpretation. If the base registers "the March riots" but not "the March clearing and injury records," or the reverse, the entry looks neutral while the frame has already taken sides. Naming disputes cannot be eliminated outright, but they can be made visible: objections to an entry's name should be attached to the entry itself, with several designations registered side by side. A base that pretends to have no naming problem is less neutral than one that openly records its naming disagreements, because the former disguises a choice as the absence of choice. A naming annotation need not await agreement to take effect: an entry bearing two designations at once reads awkwardly, yet lays the ownership of naming power out in the open. Awkwardness is a cost that can be shared; disguise is an invisible power.

When a base finds a certain class of material perpetually "below threshold," and the shape of the threshold happens to make precisely that class hardest to qualify — for instance, oral testimony and photographs held by scattered residents, which do not fit the format of official documents — the bias hides in the shape of the threshold itself. The beneficiaries are the side able to mass-produce compliant material; those who bear the cost are the people whose forms of record are at a disadvantage, whose experience, for reasons of format, never becomes "fact." Examining the structural distribution of inclusion rates across the years detects this capture better than examining any single rejection decision.

Waiting times from submission to acceptance should be tallied by the submitting faction and published. The average can show efficiency; only the distribution can show fairness: if one side's motions are accepted in ten days on average and the other's in months, then even if all are eventually accepted, the latter has already lost its timing in contention that never pauses. Time is itself part of inclusion, and delay is a form of filtering that needs to leave no trace of deletion or alteration.

The Slope of Inflation into a Monopoly on Interpretation

Inflation rarely happens in an obvious way. First a sentence appears at the end of an entry, "this record has been verified as reliable"; then, "this event was caused by external incitement"; then causal assertions arrive in the form of annotations. Each step has its convenient reason: users want conclusions directly. The criterion for recognizing the slope is whether causal or evaluative terms appear in entries that would require independent evidential support yet carry no chain of provenance; once annotations begin to carry more argument than the body text, the base is already speaking for one faction.

Citing a ready-made base is cheaper than verifying separately. This convenience has real value, and it also subjects the base to continual pressure toward expansion: users demand entries that are more numerous, more fine-grained, and closer to conclusions. Those who make the demand do not bear the cost of erroneous inclusion; that cost falls on the people later described by erroneous entries. This misalignment of convenience and responsibility is inflation's most durable engine: it requires no one to overstep intentionally, only everyone to save time.

The beneficiaries of inflation are the side close to the power of inclusion and the professional custodians: the more entries there are, the deeper every interpretation's dependence on the base, and the more irreplaceable their positions become. Those who bear the cost are the people whose records cannot enter the base, or who are fixed in place by erroneous entries. The base's periodic audits should list both rosters rather than report only overall usage; a rising usage rate may mean the base has become more trustworthy, or that contending parties have nowhere else to go.

Validity Periods and Reopening of the Factual Base

The stopping rule that Chapter 7 established for doubt becomes, here, the temporal structure of entries: every record should carry its own statement of force — the types of provenance supporting it, the known error, and the kinds of new material that would trigger review. Entries without such a statement will be cited as permanent facts, and the base's authority is precisely what cannot promise permanence; for institutions that promise permanence, Chapter 17 has already shown where such promises end: in defensive hardening.

Reopening can neither be made to wait without end nor be drowned under an immediate flood. A workable structure is this: new material is received under the same admission standard as the original entry; whether it is accepted must be answered within a limited period, and a lapse of that period counts as entry into review. The questioning faction's photographs and the committee's supplementary files travel the same path. If only one side's material is always processed in time, the path itself is already biased, and no amount of additional rule text will mean anything.

Chapter 14 showed that a correction left on the archive page is no correction. When a base entry is revised, the decisions that once cited it — compensation rulings, council minutes, public notifications — should be traceable and marked. The base therefore needs to preserve citation relations, which raises costs; without them, errors continue to live along the chains of citation, and the base gradually forfeits its standing as a public starting point, because no one can be answerable for citing a passage that is no longer accurate. Citation tracking need not exhaust every restated sentence; what it must cover are the positions where public decisions were made: if a compensation ruling rested on an entry later revised, the affected parties should be able to apply for retrial on the strength of the revision record, rather than being told that things were done at the time according to "common facts" and that no retroactive inquiry will be entertained.

Entries past their validity period and unsupported by new material should pass, marked as lapsed, into the base's archival layer rather than disappear. Deletion makes two questions indistinguishable: has this entry expired, or was it quietly pulled down? Lapsed entries, together with their reasons for lapsing, are themselves part of the base's history and a sample for testing whether the rules of inclusion are actually carried out in practice. A base that leaves no corpses cannot prove to anyone that it has killed no one.

Unilateral Maintenance after One Side Withdraws

If the committee faction announces that it will no longer take part in the base, does it still mean anything for the questioning faction to keep registering entries? It does, but the nature changes: a unilateral base is only one side's archival organization and no longer automatically holds the standing of a common starting point. To keep registering under common rules — including material unfavorable to itself — is its only route back to that standing; once it shifts to filtering by the logic of propaganda, it remains one side's archive forever.

The credibility of a unilateral base is measured by how it handles unfavorable material: whether it includes the contradictions between its own side's testimony and the official record, and whether it preserves the objections, and the partial admissions, that the other side raised before withdrawing. A unilateral archive that admits only favorable entries, however rigorously its rules are written, is propaganda; and an honest unilateral archive stands closer to public fact than a captured "common" base.

When the two factions make contact again, the cross-check list between the unilateral base and the other side's archive is the first material for restoring a common base: which entries both sides hold, which only one side holds, and which stand in direct contradiction. Commonality restarts from the overlaps, not from apology or general reckoning; the very existence of the cross-check list means that reunion need not presuppose one side's total confession of error, which precisely lowers the threshold of reunion.

Contention When the Common Surface Is Minimal

When the two factions cannot agree even on what counts as a record, contention moves up to the meta-level: thresholds, naming, format. This is not necessarily failure: contention at the meta-level still takes the base as its object and is more corrigible than mutual accusation with no object at all. The danger is that meta-disputes are used to postpone all registration indefinitely; these too need a stopping rule: which disagreements over rules can coexist — two formats annotated in parallel, for instance — and which must be settled first — the minimal definition of source independence, for instance. Contention at the meta-level has one further benefit that is easy to overlook: in discussing "what counts as a record," both sides rehearse in advance a way of challenging material rather than persons. That rehearsal carries back into contention over specific entries, so that the base's everyday revisions do not escalate every time into factional showdowns.

On entries that admit of no agreement, the base can register the disagreement itself: one side claims such-and-such with material attached, the other claims the opposite with material attached, and the contradiction remains unresolved. This looks like concession, but it turns disagreement from a war of attrition in words into an object that has a location, has material, and is open to review. When new material arrives, having preserved the full issue and its grounds can reduce the preparatory work of renewed verification, but it cannot guarantee that these entries are resolved first.

Disagreements over values, claims for compensation, and historical judgment still proceed in deliberative venues outside the base. The base's success is not measured by whether these disputes are quelled, but by whether common material remains available while they run. Those who expect the base to end all conflict will pursue, through ever-expanding inclusion, what it cannot do — and this very expectation is the engine that a monopoly on interpretation most gladly borrows.

The Failure Boundaries of the Common Base

In some conflicts the core lies not in facts but in the ordering of values, or the two sides have long agreed on the facts and simply cannot accept what they mean. For such disputes, forcibly expanding the base merely translates value conflict into factual dispute and manufactures a false appearance of resolvability. The way to recognize this situation is to look at the structure of citation: when the facts cited by both sides largely coincide and the dispute persists, adding further entries should stop; the problem is not in the material but elsewhere.

The signs of capture can be observed: passage rates of motions to add or delete entries distributed by faction over long periods; one side's material taking markedly longer to be accepted; causal language without provenance beginning to appear in entry annotations; "not accepting the base" becoming grounds for sanction; custodial roles concentrated in one faction and its affiliates. Any single sign can admit another explanation; when several coexist, those who keep citing the base should realize that what they are citing has already become a weapon.

Recovery does not require destroying the base and starting over. One can begin by restoring the records of shelved motions, publishing the distribution of acceptance periods and passage rates, stripping the interpretive annotations that have crept into entries, and re-electing custodial roles so that they are dispersed. Harder than rebuilding the structure is restoring the habit of citation: the side that has suffered capture needs time to test the new procedures, and during that period, registering entries of disagreement and tolerating dual-track citation are the practical ways to rebuild trust — declarations and apologies cannot substitute for this observable stretch of time.

From Minimal Starting Point Back to a Contestable Common Ground

Since minimal shared facts constitute a public standard, they inherit the requirement stated at the end of Chapter 18: if a standard accepts argumentation only at its founding, it will continue to govern later subjects after circumstances have changed. Every part of the base — entries, thresholds, naming, custody, exclusion — should therefore carry its own validity period and conditions for reopening, keeping itself always an object that can be contested rather than a premise of contention tacitly conceded. Qingya Valley does not need to choose between "common facts" and "each our own truth." It can start from overlaps too small to shrink any further: a date, a contract total, a water-level chart both factions have cited, while keeping the parallel archives, registering disagreements, auditing the distribution of inclusion, and subjecting every exclusion and every act of naming to review. The smaller and truer the base, the wider the field open to interpretation; the fatter and harder the base, the faster contention degenerates into a struggle over the base itself. The boundary is thus clear: without a base, contention degenerates into a pure contest of force, and the weakest go out first; when the base inflates into a monopoly on interpretation, it turns from common starting point back into one side's position, ultimately forcing the other side to build a parallel base of its own, and the facts split apart again. Letting contention share minimal facts is not a search for a neutral high ground that needs no maintenance; it is the making of "what at least we are all talking about" itself into a public work that demands continuous upkeep and can be reopened at any time.

The next chapter takes up the multiple gateways to preserved records: when archives beyond the common base also need to be found, understood, and reviewed, how the number and quality of gateways determines the effective reach of public memory.