Montpellier Opens the Boxes

The verb to open

“Montpellier opens the boxes” compresses years of work into an active verb. The boxes did not open themselves, and the first useful act was not spectacular. The university had to receive a deposit, establish responsibility for the objects, inspect their condition, identify groupings, preserve fragile sheets, decide how to number them, and create surrogates that could be served without repeated handling of the originals. Only then could a web address give the appearance of immediate access. 11Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”; Institut des Hautes Études Scientifiques, “Some of Grothendieck’s Archives Published Online by the University of Montpellier”.

That sequence corrects the mythology of the treasure chest. A treasure chest holds objects whose value is evident once the lid rises. An archive holds relations whose value depends on provenance and order. A loose page of symbols can be arresting; a dated sequence of fifty pages, tied to a course, a letter, and a later typescript, can be historically intelligible. The difference is produced by archival description and mathematical comparison. To publicize a page while severing those relations may make it more visible and less usable at the same time.

The Montpellier papers arrived through Jean Malgoire, a former student and university mathematician. The University of Montpellier’s May 10, 2017 article says Grothendieck entrusted the archive to him in 1990; IHES says the transfer occurred in two stages, in 1990 and 1995. Both accounts place Malgoire’s university deposit in 2010. Without the transfer instrument, the earlier date remains a reported claim rather than a settled chronology. Provenance can be refined only by comparing deeds, receipts, inventories, and participants’ accounts. 22Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”; Institut des Hautes Études Scientifiques, “Some of Grothendieck’s Archives Published Online by the University of Montpellier”.

The date matters because the material itself covers many phases. The archive is described as representing most of Grothendieck’s mathematical papers from 1949 to 1991, not simply his Montpellier years. It reaches back to the period in which functional analysis and algebraic geometry transformed his career and forward to the eve of his removal to Lasserre. A deposit made near the end of that range can preserve retrospection, rearrangement, and reuse as well as original chronology. A biography must not assume that the order in which pages reached Malgoire is the order in which they were written. 33Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”; Institut des Hautes Études Scientifiques, “Some of Grothendieck’s Archives Published Online by the University of Montpellier”; Réseau national des bibliothèques de mathématiques, “Archives mathématiques d’Alexandre Grothendieck: 18 000 documents en libre accès”.

From deposit to pixels

The university’s own account names three operations before public access: inventory, conservation, and digitization. Digitization began in 2016; a substantial online portion opened on 10 May 2017. That timeline is brief by archival standards and long enough to defeat the fantasy of an unmediated dump. Someone determined resolution, color, recto and verso, file sequence, identifiers, thumbnails, and the relation between a catalog description and an image. Each choice is technical and interpretive. 44Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”; Université de Montpellier, “Grothendieck: 18,000 Pages Now Available”.

At launch, official notices repeated two approximate quantities: about 28,000 pages in the corpus and about 18,000 accessible online. Those numbers should remain attached to the date and the institutions that gave them. They are not permanent counters. New description can change what counts as a page, a blank verso, a duplicate, an enclosure, or a restricted item. Nor does the difference of roughly 10,000 identify a single category that may safely be called “secret.” Letters can involve the rights and privacy of correspondents; some material may require further processing; the exact composition of the unavailable portion must be taken from the archive’s current documentation, not deduced by subtraction. 55Institut des Hautes Études Scientifiques, “Some of Grothendieck’s Archives Published Online by the University of Montpellier”; Réseau national des bibliothèques de mathématiques, “Archives mathématiques d’Alexandre Grothendieck: 18 000 documents en libre accès”; Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”.

The archive description lists handwritten notes, typescripts, occasional printed material, and letters. These forms encode stages of work. Handwriting may be a first exploration or a copied fair version. A typescript may look finished yet carry extensive correction. A printed offprint can be annotated until it functions as a new document. A letter can contain a theorem, a request, a complaint, or all three, and it remains jointly entangled with a recipient’s history. “Manuscript” is a useful umbrella only if the distinctions beneath it are restored when claims are made. 66Institut des Hautes Études Scientifiques, “Some of Grothendieck’s Archives Published Online by the University of Montpellier”; Réseau national des bibliothèques de mathématiques, “Archives mathématiques d’Alexandre Grothendieck: 18 000 documents en libre accès”.

The portal says that the classification is substantially faithful to Grothendieck’s own arrangement where that arrangement existed. This is not a minor archival courtesy. Mathematical thought often proceeds by dossiers: a theme accumulates examples, failed definitions, diagrams, cross-references, and provisional names before acquiring a linear exposition. Preserving the dossier can reveal which problem generated a definition and which definition was imposed retrospectively. But fidelity to original order must not become a cult. The author may have changed a folder for convenience; an assistant may have inserted an offprint; a later move may have broken the sequence. The archive’s phrase “when it existed” is the limit. 77Réseau national des bibliothèques de mathématiques, “Archives mathématiques d’Alexandre Grothendieck: 18 000 documents en libre accès”.

Digital access adds a second order over the first. A web interface can sort by title, date, or cote; a search result can jump from one decade to another. That power is one of digitization’s gifts, especially to researchers far from Montpellier. It can also make adjacent pages seem independent and give an OCR hit more authority than the barely legible line from which it came. The responsible reader records the stable identifier, page image, surrounding leaves, catalog description, access date, and transcription uncertainty. A bare browser link is a route, not a citation.

What a mathematical archive can prove

Suppose a leaf contains a definition recognizably similar to a notion published years later by another mathematician. The image alone does not prove priority. The date must be secure; the terminology must be interpreted in its contemporary context; the definition’s scope and consequences must be compared; evidence of circulation must be sought; independent routes to the later result must be considered. A private anticipation can be important without making the later author a borrower. Conversely, the fact that a note was unpublished does not make it historically inert if letters, seminars, or students transmitted the idea.

The archive can support several different claims, and they should not be collapsed. It may show that Grothendieck considered a question by a certain date. It may show that he formulated a definition or conjecture. It may show sustained work through revision. It may show communication to a named recipient. It may, with enough mathematical checking, establish that an argument was complete. It almost never shows, by itself, what everyone else knew or why another person acted.

This layered approach is particularly necessary for papers after 1970. The old story that Grothendieck simply abandoned mathematics is contradicted by the quantity and substance of La Longue Marche à travers la théorie de Galois, Pursuing Stacks, Esquisse d’un Programme, and other notes. The contrary story—that every later page belongs to one fully formed hidden system—is no better. The Montpellier archive makes activity visible; it does not make the activity uniform. The years include teaching, programs, long investigations, autobiographical reflection, and works with different intended audiences. 88Sébastien Maronne and Bertrand Toën, “Préface,” Grothendieck mathématicien: Le temps des réflexions 1973–1991 (exposition), November 2024, https://exposition-grothendieck.org/preface/; Grothendieck, “Esquisse d’un Programme,” 7–48; Grothendieck, “Esquisse d’un Programme,” 003; Grothendieck, À la Poursuite des Champs (Pursuing Stacks).

The curators of the later exhibition chose exactly this problem as their subject. Sébastien Maronne and Bertrand Toën called their project Grothendieck Mathématicien — Le temps des réflexions 1973-1991. Their preface explicitly turns away from a picturesque great-man biography built from ecology, meditation, and disappearance. Instead, it asks what the thousands of pages reveal about continuous mathematical activity after the move to Montpellier. That is an argument made through selection, not a declaration that biography is irrelevant. It restores a neglected mathematical chronology by bracketing, for twelve panels, other chronologies that remain real. 99Maronne and Toën, “Préface”.

Twelve panels from eighteen thousand pages

The exhibition first took material form in Toulouse, where the BnF catalog records a showing from 9 December 2024 to 31 January 2025. Its accompanying booklet is thirty-two pages, authored by Maronne and Toën with collaboration by Damien Calaque. Philomène Longchamp created the graphic conception and illustrations. The project describes itself as a mobile exhibition of twelve panels intended chiefly for a university audience, available for loan in digital high definition or physical form. 1010Maronne and Toën, Le temps des réflexions; “Grothendieck mathématicien: Le temps des réflexions 1973–1991 (exposition),” https://exposition-grothendieck.org/; Bibliothèque nationale de France and Société mathématique de France, “Alexandre Grothendieck, mathématicien et militant”.

Those proportions reveal the discipline of exhibition making. Eighteen thousand online pages cannot be miniaturized into twelve walls. A panel must select a text, crop or reproduce an image, supply context, and choose a path that a standing reader can follow. The booklet can restore bibliography and detail, but it too is finite. The result is not “the archive online” in another format. It is a scholarly essay made spatial, with authors whose decisions should be visible and citable.

The exhibition’s path follows several late mathematical works as samples of a broader enterprise: La Longue Marche in 1981, Pursuing Stacks in 1983, Esquisse d’un Programme in 1984, and Récoltes et Semailles from 1984 to 1986. Its claim is that these writings contradict the picture of a long mathematical afterlife reduced to the tail of a brilliant earlier comet. The preface also resists the opposite simplification. It acknowledges the difficulty of communicating serious mathematics and prefers a complex analysis to formulas made simple enough to deceive. 1111Maronne and Toën, “Préface”.

In January 2025 a version appeared at the BnF study day. The location joined the Montpellier scans to the newly accessioned Paris manuscripts without confusing their custody. The foyer panels treated work from 1973 to 1991; the day’s talks reached activism, teaching, film, autobiography, and late reflections on evil. The archive thereby generated not one memorial but a set of encounters among document types and disciplines. 1212Bibliothèque nationale de France and Société mathématique de France, “Alexandre Grothendieck, mathématicien et militant”; Bibliothèque nationale de France and Société mathématique de France, “Les mille et une pages mathématiques d’Alexandre Grothendieck”.

The exhibition continued to travel. The ENS-PSL Mathematics and Computer Science Library records a showing from April 8 through July 8, 2026 and names the creators and their institutional affiliations. That is the latest venue verified for this account, not a final venue. A mobile exhibition has a changing itinerary, and both loan conditions and online resources can be revised. Every exhibition-status statement therefore carries a verification date. 1313École normale supérieure – PSL, “Exposition: Grothendieck, le temps des réflexions, 1973–1991,” 8, 2026.

An online exhibition is not an image license

The exhibition exists across several digital layers. Its website supplies a preface, a poster, and loan information. A booklet record sits in HAL and is marked open access. The BnF catalog supplies a bibliographic description. Venue pages quote the curatorial text. These layers make the project discoverable, but none alone grants a future biographer permission to lift every manuscript detail or illustration.

The poster, for example, is credited to Philomène Longchamp. Open display of the JPEG is evidence of creator and context, not of a reusable license. Manuscript pages within panels can involve Grothendieck’s authorial rights, the university’s digital surrogate, the graphic designer’s arrangement, and perhaps third-party material. 1414“Grothendieck mathématicien: Le temps des réflexions 1973–1991 (exposition)”.

The portal as a scholarly instrument

A digital archive succeeds when readers can recover context rather than images alone. Stable cotes permit a mathematician in one country and an archivist in another to discuss the same leaf. Page sequence permits an assertion to be checked against correction. High-resolution recto and verso preserve evidence that a cropped quotation could hide. Descriptive notes can warn of uncertain dates or composite files. In the most useful case, the portal becomes a distributed reading room rather than a picture gallery.

That instrument remains vulnerable. Domains change, certificates expire, viewers are replaced, and links into a proprietary interface break. On August 22, 2026, direct retrieval of the Montpellier domain failed in the production environment through a TLS or DNS transport error. Institutional pages from 2025 and 2026 nevertheless continued to direct readers to the scans. These observations establish neither a global outage nor uninterrupted availability; they support only a local access statement dated to the time of writing. 1515Université de Montpellier, “Archives Grothendieck”; Hartemann et al., “Les mille et une pages de Grothendieck”; Bibliothèque nationale de France and Société mathématique de France, “Alexandre Grothendieck, mathématicien et militant”.

The chronology therefore retains two explicit limits: the entrustment date remains contested, while live portal availability is verified only locally as of August 22, 2026 for this account.

That mundane observation belongs in a biography of the digital afterlife. A statement such as “the complete archive is freely available forever” is technologically naïve and factually wrong twice: the launch itself described only a portion, and no web service carries a guarantee of permanence. Better practice records an institutional landing page, stable item identifiers when obtainable, access dates, local copies only when law and policy allow them, and checksums for files lawfully incorporated into a project. If an interface changes, the citation can still point back through catalog data.

Access can also change for legitimate reasons. A repository may discover personal data, disputed rights, preservation risks, or inaccurate metadata. It may add material after clearance. Such changes are not necessarily censorship or progress; their meaning depends on documented reasons. The SAA code’s balance is useful even outside its jurisdiction: maximize access, respect legal and ethical restrictions, document decisions, and state why a restriction exists and for how long when that can be known. 1616Society of American Archivists, “SAA Core Values Statement and Code of Ethics”, Judgment, Privacy.

Reading the boxes without worshipping them

The archive’s scale encourages two kinds of awe. The first worships abundance: surely tens of thousands of pages contain an undiscovered revolution. The second despairs of abundance: surely no one can read enough to say anything. Both responses evade the unit of work. A research question narrows a series; a series leads to dates and cotes; a page leads to neighboring pages; a mathematical claim leads to published references and living specialists. The archive becomes readable by accountable reduction, not by pretending the whole has been mastered.

Malgoire’s image of seeds is strongest when detached from prophecy. A seed may germinate, remain dormant, or fail in a given soil. Cultivation changes what survives. In mathematical history, cultivation means transcription, seminars, proof checking, comparison, and credit. The person who reads an unfinished argument and makes it rigorous performs new work; the archive should make both the inherited suggestion and the later construction visible. 1717Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”.

The boxes also correct a vocabulary of solitude. Grothendieck wrote many pages alone, but their survival involved a former student, university staff, digital technicians, mathematicians, designers, libraries, and audiences. The exhibition’s twelve panels are collective products about a man often cast as uniquely isolated. That does not erase withdrawal. It shows that posthumous access is an institutionally made relation between solitude and community.

Montpellier opened the boxes in this procedural sense: it accepted custody, preserved order where order existed, created an inventory, digitized a substantial portion, and built a route for readers. Maronne, Toën, Longchamp, and their hosts then opened a route through the route, selecting a mathematical itinerary for public view. Neither opening is final. The archive can be rearranged in scholarship without rearranging its physical leaves; the exhibition can travel; rights can be clarified; interpretations can be challenged.

The most important result is not that the papers have yielded a secret. It is that questions can now be posed with better evidence: when a theme entered the notebooks, how a lecture became a program, which late concepts circulated, how expository style changed, and where a claim remained conjectural. The next chapter follows several answers already visible in the boxes and in the working mathematics of other people. The archive gives those paths documentary depth. It does not own their destinations.

Recto, verso, sequence

The smallest archival decisions can alter a mathematical reading. A page numbered by Grothendieck may carry calculations on its reverse that belong to another project. An archivist must decide whether the verso follows the recto in image sequence, whether it receives its own page label, and how a reader learns that the relation may be merely physical. If the sheet was reused, two chronologies can occupy one object.

Margins matter for the same reason. A main text can be early and a marginal correction late. Arrows can redirect paragraph order. A different ink may mark revision, dictation, or nothing chronologically significant; only comparison supports an inference. Cropping to the written block produces a cleaner display and can remove punch holes, paper changes, old folder numbers, and notes that establish provenance. The full archival image and the designed detail serve different purposes and should not be confused.

Typescripts create another trap. Their regularity suggests a settled edition, but Grothendieck’s typescripts can be corrected, augmented, and reorganized. The historian should record the base typing, handwritten layers, replacement leaves, and pagination. A transcription that silently incorporates every correction may be readable while destroying evidence of sequence. A diplomatic transcription preserves more of the physical state but can overwhelm a general reader. A serious edition can provide both levels and explain the relation.

Digital presentation should preserve these distinctions in metadata. “Page 37” is ambiguous unless readers know whether 37 is an authorial number, an archival folio, or the viewer’s image count. Stable cotes must survive interface redesign. Downloaded filenames should not be the only identifiers, because filenames can change. The notes therefore cite collection descriptions rather than pretend to item-level references that could not be recovered.

This attention to sequence is not antiquarianism. A theorem’s status can turn on whether a counterexample was written before or after it. A program’s influence can turn on whether a typescript was mailed. A change from “there exists” to “one might seek” can separate assertion from hope. The marks that establish those facts live at page level.

From archive order to exhibition order

An archive and an exhibition answer opposite spatial problems. The archive aims to preserve relations among all retained materials and to make retrieval repeatable. The exhibition must break that scale, bringing a few objects into a path that can be grasped during a visit. Archive order resists a single story; exhibition order makes one story explicit.

Maronne and Toën chose a chronological and thematic itinerary through Grothendieck’s mathematics from 1973 to 1991. The dates begin with his arrival as a professor at Montpellier and end with his move to Lasserre. Within that interval, the curators focus on works that make continuing mathematical reflection visible. Their preface identifies the Montpellier scans as the principal documentary and iconographic source. 1818Maronne and Toën, “Préface”.

That declaration performs valuable source work. It tells viewers that images did not float in from an anonymous web search and that the exhibition’s argument arose from a named corpus. It also permits critique. A reader can ask whether the selected pages represent the works, whether the 1973 boundary suppresses earlier continuity, or whether the mathematical focus underplays teaching and activism. Explicit selection creates the possibility of an informed alternative.

The twelve-panel form requires compression at every level. An eight-hundred-page handwritten investigation may receive one image and a paragraph. Mathematical notation must remain legible at viewing distance. French prose needs a reader’s route. The graphic designer controls hierarchy, color, and juxtaposition, influencing whether a manuscript appears as evidence, texture, or icon. Naming Philomène Longchamp is therefore not a decorative credit; design belongs to the exhibition’s argument. 1919Bibliothèque nationale de France and Société mathématique de France, “Alexandre Grothendieck, mathématicien et militant”; Maronne and Toën, Le temps des réflexions; Maronne and Toën, “Préface”.

The booklet expands but does not erase the panels’ constraints. Its thirty-two pages supply a portable record and bibliography. The HAL deposit broadens access to that record, while the BnF catalog anchors authorship and publication data. A future biography should cite the booklet when borrowing an interpretation and the original archive when making an item-level claim. Curatorial scholarship is a source, not a substitute for the source it interprets. 2020Maronne and Toën, Le temps des réflexions.

The reader standing before a panel

The exhibition addresses a university audience, and its preface acknowledges the difficulty of showing serious mathematics. That choice resists a common hierarchy in which accessibility means removing definitions until only personality remains. It assumes that respect can take the form of asking effort from viewers while giving them enough structure to begin. 2121Maronne and Toën, “Préface”; “Grothendieck mathématicien: Le temps des réflexions 1973–1991 (exposition)”.

There are several possible readers in front of the same panel. A mathematician may recognize a moduli space and ignore handwriting. A historian may examine dates and corrections while treating the formula cautiously. A student may encounter an unfamiliar program through a diagram. A designer may notice how the page has been cropped. A relative or former colleague may read a name as part of a personal history. No panel controls all those encounters.

An online version changes the reader again. Zoom makes a formula examinable but removes physical scale. Hyperlinks can supply full references or invite distraction. A screen reader needs meaningful alt text that cannot stop at “old manuscript.” French handwriting may not be machine-readable. High contrast helps some readers and can misrepresent ink. Accessibility is thus an archival and mathematical problem as well as a design requirement.

A manuscript caption should identify the object, state when and where it was made within the limits of the catalog, explain why that leaf is shown, and direct attention to the relevant feature. A caption offering praise alone meets none of those requirements. A caption that identifies a draft from Esquisse d’un Programme, a revised phrase, a repository, and a cote turns an image into evidence.

Alt text should not attempt to transcribe an entire dense leaf. It should identify layout and the feature central to the surrounding argument, while a separate transcription supplies text if permitted. A decorative manuscript texture has little justification when it cannot be read. An exhibition or edition should favor fewer, fully identified leaves over a montage that converts work into atmosphere.

The unavailable remainder

The difference between 28,000 and 18,000 invites speculation. Some public summaries associate unavailable material with correspondence and rights. That is plausible and supported at a general level by descriptions of letters in the fonds, but no itemized restriction schedule has been recovered. It therefore does not write, “The remaining ten thousand pages are private letters.” 2222Institut des Hautes Études Scientifiques, “Some of Grothendieck’s Archives Published Online by the University of Montpellier”; Réseau national des bibliothèques de mathématiques, “Archives mathématiques d’Alexandre Grothendieck: 18 000 documents en libre accès”.

This refusal illustrates a wider rule. A numerical remainder is not a category. Pages may be awaiting processing, duplicate, blank, withheld for legal reasons, technically unavailable, or counted under a different convention. Only a current institutional statement can classify them. Archive status evolves as rights are cleared, descriptions improve, and conservation proceeds.

Restricted access can preserve future access. A fragile original may be withheld while a surrogate is prepared. A letter may remain closed during a living person’s lifetime. A repository may require supervised consultation. Those conditions can frustrate research without constituting permanent suppression. The reason and expected duration, where public, are part of the fact.

Opacity about restriction can nevertheless damage trust. Researchers need to know whether a gap reflects missing material, an embargo, processing, or a broken link. The SAA code recommends transparent, documented, and equitably applied restrictions. Again, that code does not govern Montpellier; it articulates a professional ideal against which any archive description can be read. 2323Society of American Archivists, “SAA Core Values Statement and Code of Ethics”, Judgment, Privacy.

The biography should mirror that transparency. When no image was accessible, it says so. When a fact comes from a university announcement rather than the item, it says so. When a page count is approximate, it preserves the approximation. These small qualifications prevent absence from being filled by confidence.

Discoveries and negative results

Archive publicity favors discoveries. A previously unknown theorem, letter, or prediction provides a headline. Most sustained archival work produces humbler results: a date corrected, an attribution narrowed, a draft connected to a lecture, a supposed novelty recognized as copied notes, an illegible word left unresolved. Negative results are part of knowledge.

For mathematics, establishing that a page does not contain a complete proof can be valuable. It protects later authors from anachronistic credit claims and shows where Grothendieck saw the obstacle. Establishing that a concept did not circulate can distinguish private anticipation from influence. Establishing that two files once thought sequential are independent can change the history of a program.

The exhibition’s focus on samples makes no claim to excavate a lost theorem. It argues for continuity of mathematical work by presenting dated traces and texts with known afterlives. That is already a historical correction to the absolute language of abandonment. It leaves room for unevenness: a person can withdraw from institutions, cease formal publication, and continue mathematical research without all three processes sharing one date.

This is why the boxes open repeatedly. They opened physically for preservation. They opened digitally for distributed consultation. They opened curatorially in panels and a booklet. They open mathematically whenever a reader reconstructs an argument and submits it to proof. They open historically when another archive supplies the missing side of a correspondence.

No opening is ownership. A curator’s route can be revised, a mathematician’s interpretation refuted, and a portal redesigned. The collection remains the reference point against which those acts are checked. Its value lies not in promising a final surprise but in sustaining correction over time.

The labor behind “free access”

The phrase libre accès describes the user’s lack of a paywall or physical journey; it can obscure the cost borne elsewhere. Storage, scanning, conservation, metadata, web maintenance, and staff time require institutional resources. The former student who preserved boxes at home also performed labor before institutional processing began. Public access rests on those investments. 2424Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”; Université de Montpellier, “Grothendieck: 18,000 Pages Now Available”.

Recognizing labor changes the rhetoric of gift. Grothendieck created and entrusted the papers. Malgoire preserved and deposited them. The university transformed them into a service. Curators interpreted them. Readers benefit without any one participant becoming sole author of the archive’s afterlife. Credit should follow tasks.

Sustainable access requires maintenance beyond launch day. Checksums detect file corruption. Redundant storage protects masters. Migration keeps formats readable. A contact route allows errors in metadata to be reported. A preservation copy and a delivery copy serve different needs. None of these operations appears in the luminous image of a manuscript on a screen, yet each determines whether it will remain.

The archive’s online future cannot be guaranteed by this biography, but the biography can avoid making it more fragile. It can cite institutional identifiers, retain access dates, avoid deep links when stable cotes are unavailable, and record hashes for any lawfully received image file. It can refrain from mirroring protected material merely to insure against a broken website. Preservation is an institutional and legal relation, not an excuse for unauthorized copying.

The boxes, then, open into infrastructure. That conclusion is deliberately less romantic than treasure. Infrastructure permits many discoveries, including the discovery that an expected treasure is not there. It distributes access, carries correction, and makes later paths possible. From that base, the next chapter can follow mathematics that lives not as paper alone but as definitions, theorems, programs, and arguments in other hands.

A method for the first reader

The first reader of an unfamiliar dossier should begin by resisting transcription. Before converting handwriting into searchable text, the reader records the cote, container, leaf range, dimensions, pagination, visible dates, titles, writing instruments, and relation to enclosures. Photographs or scans are checked for missing edges and reverses. Existing catalog language is copied as catalog language, not silently adopted as the author’s title.

Only then does transcription begin. Uncertain letters receive brackets or another declared notation. Mathematical symbols are not modernized without a parallel diplomatic version. Line breaks matter when they distinguish a displayed claim from a marginal afterthought. A second reader checks the result against the image. If the archive restricts reproduction, the research notes record what verification remains possible for others.

Mathematical reconstruction is a separate document. It can introduce modern notation, supply definitions, and reorder steps, but every intervention should point back to the diplomatic text. Missing arguments are labeled missing. A plausible lemma is not inserted under Grothendieck’s name merely because it would make the proof work. If later mathematicians complete the argument, the completed proof has joint historical layers.

Publication claims wait for comparison. Search begins with Grothendieck’s published works, known circulating manuscripts, correspondence, seminar literature, and later papers. Terminological difference can conceal identity; identical terminology can conceal changed definitions. Specialists in the precise field should review novelty before a press announcement.

Finally, the reader asks whether public quotation or facsimile is necessary. An archive finding aid can often support provenance; a paraphrase can report subject; a short quotation can establish distinctive wording. Full leaves are reserved for cases in which handwriting, correction, diagram, or layout carries the argument. Rights have to be settled before design, not after a page proof has made an image seem indispensable.

This protocol slows the first claim and accelerates later trust. It allows readers who never touch the physical box to distinguish what the manuscript says from what its interpreter supplied. Opening becomes reproducible work rather than privileged encounter.

The protocol should include failure. If handwriting cannot be resolved, the transcription marks it. If two leaves cannot be ordered, both possibilities remain. If a formula cannot be verified, the finding is historical rather than theorematic. If rights prevent an image, the catalog reference can still guide qualified readers. Research becomes less credible when every obstacle disappears from its final prose.

Montpellier’s achievement is therefore measured not only by quantity but by the possibility of audit. Page images, order, identifiers, and institutional custody let later readers revisit claims. The exhibition adds an authored interpretation whose sources are named. Together they replace the sealed-box legend with a chain of inspectable choices, which is the more durable kind of opening.

A good portal also permits correction from outside the repository. Researchers can report a reversed image, uncertain date, or broken sequence, while archivists retain authority over the catalog. That exchange turns remote access into collaboration without surrendering provenance. The useful measure is not whether every user agrees with the description, but whether disagreement can identify the same object, present evidence, and leave a documented decision for the next reader. Transparent change logs can show whether the repository corrected metadata, altered access, replaced a scan, or merely repaired the interface for users that day.

Such transparency makes later correction possible, visible.