The Past Does Not Survive on Its Own. Someone Decides What Is Kept.

After 1949 a great many Republican-era political and military figures were stranded in the United States with no archives in hand — the archives sat locked in Taipei and Beijing, each serving its own official war narrative. Around 1957 Columbia University set up a Chinese Oral History Project and hired a doctoral student, Tang Te-kang (1920–2009). He recorded Hu Shih first, then took on Acting President Li Zongren (1891–1969).

The interviews ran intermittently for seven years, and the trouble surfaced fast: Li’s recollections did not match the documents — dates drifted, credit was shaded, and his judgements of Chiang Kai-shek carried old grievances. Tang neither transcribed blindly nor ghost-wrote a smoother version. Instead he checked each claim against newspapers, telegrams and government gazettes, and set the discrepancies alongside the text as annotations; his own description of the craft was “three parts testimony, seven parts verification.” The Memoirs of Li Zongren thus became a two-layer document: a participant’s memory, plus a running audit of that memory. On 20 July 1965 Li returned to Beijing; the manuscript had already taken shape in New York, and survived intact for exactly that reason.

Incident postmortems and exit interviews work the same way: ask through a direct manager and you get a safe script; ask through a third party, with statements laid beside the log timeline, and you may recover the real trade-off. The value of a record lies less in what was said than in whether it sits somewhere it can be cross-checked.

Postwar Japan kept its textbook certification system: books are written privately but must pass Ministry of Education screening before entering schools. New Japanese History, by the Tokyo University of Education scholar Ienaga Saburō (1913–2002), was failed in the 1962 screening round and passed the next year with some three hundred required revisions, touching the Nanjing Massacre and the treatment of Unit 731.

On 12 June 1965 Ienaga sued the state for damages, invoking Article 21 of the constitution and its ban on censorship, thereby lifting the question from “is this claim historically sound” to “may the state screen a historical account before publication at all”. On 17 July 1970 the Sugimoto ruling of the Tokyo District Court sided with him: certification amounted in fact to prior censorship, and was unlawful — a ruling promptly overturned on appeal. Three suits followed, spanning thirty-two years. On 29 August 1997 the Supreme Court issued its final judgement: most claims were dismissed, but the demand to delete the passage on Unit 731 was held unlawful. Thirty-two years bought one finding of overreach — and one paragraph that could no longer be removed.

“Prior review versus later accountability” recurs in every medium: several U.S. states now legislate what history may be taught, and platform governance swings between pre-emptive blocking and after-the-fact liability. Choose prior review and you gain efficiency while the reviewer owns the boundary alone; choose later accountability and you pay more, but errors can be corrected from outside.

In 1979 the veteran Jan Scruggs launched a drive for a memorial to the Vietnam dead. The war offered no victory to celebrate and the official narrative was still torn. The organisers set one decisive rule: the design competition would be judged anonymously, and entries could carry no political statement. Of 1,421 submissions received in 1981, the jury chose entry 1026 — by Maya Lin, a 21-year-old undergraduate in Yale’s architecture programme.

The design was two walls of polished black granite cut into the ground in a V, carrying more than fifty-seven thousand names. Two decisions rewrote the grammar of commemoration: the names run in order of date of death rather than rank and alphabet, so men killed in the same ambush stand together and the visitor reads events rather than a roster; and the stone is polished to a mirror, so reading the names means seeing your own face among them. The backlash followed at once: the veteran Tom Carhart called it “a black gash of shame,” and Interior Secretary James Watt briefly withheld the construction permit. Work began on a compromise: the wall was built and dedicated on 13 November 1982, with Frederick Hart’s realist bronze Three Soldiers and a flag added to one side. Two grammars of memory sharing one lawn is itself an accurate footnote to that war.

The fights after 2020 over removing Confederate statues are the same argument: a monument does not record the past, it declares a present position, which is why every generation refights it. Teams face the same choice — does the wall carry only the wins, or does it list the failures?

From December 1989 crowds across East Germany stormed local State Security (Stasi) offices to stop the destruction; on 15 January 1990 the Normannenstraße headquarters in Berlin was occupied. The inventory: roughly 111 kilometres of files, more than 174,000 unofficial informants (IM), and the paper that had not been shredded in time, stuffed into some 16,000 sacks. Many argued at the time for sealing or even destroying the lot, to keep reckoning from spiralling out of control.

In December 1991 the Stasi Records Act took effect and Joachim Gauck, a former pastor, became the first Federal Commissioner for the files. The law made an unusual choice: the archive went neither to prosecutors nor to the public, but first back to the people who had been watched — you could apply to read your own file and see who had been reporting on you. Individual access requests have passed seven million over three decades. Timothy Garton Ash recorded in The File (1997) how he went door to door to meet his own informants: the motives were mostly banal — not fanaticism, but timidity, vanity and small conveniences. Spain went the other way: the 1977 Amnesty Law and an unspoken “pact of forgetting” sealed the Franco era, at the cost of having to reopen it thirty years later under the 2007 Historical Memory Law.

“How long do we keep it, and who may read it” is now a decision every company makes: the GDPR right to be forgotten (Google Spain, 2014) and a log-retention policy are two sides of one coin. Retain by default and you pre-assemble a dossier for whoever holds power later; delete on expiry and you give up the ability to clear your own name. A retention period is a political decision, not a technical parameter.