About

Who is speaking

Nizar Yaïche, former Minister of Finance of the Republic of Tunisia, independent researcher and practitioner. The book is written in that capacity, outside any official mandate: it commits no institution, and its author draws no regulatory authority from it.

The author’s note, in the book.

Legal notice

Publication director

Nizar Yaïche

Contact

nizar.yaiche@ystrat.com

Licence

Creative Commons CC BY 4.0 — reproduction and adaptation are free, provided the author and the book are cited.

Nature of the site

Personal research site, with no advertising, no tracker and no user account.

Personal data

No collection: no account, no analytics cookie, no form. The only possible transmission to a third party is the question put to the assistant, described on the page use of artificial intelligence.

What the site keeps in your browser

Strictly what reading comfort requires, held in the tab’s memory (“sessionStorage”): the records you have opened in the diagram, and the last paragraph reached in a page of the book. This information never leaves your device and disappears when the tab is closed. No cookie is set.

Where this publication stands

Version 2.3 of the text (27 July 2026), open for comment. It has not been submitted to peer review. It is self-published, without a publisher, and deposited in open access on Zenodo on 14 September 2026 under the identifier DOI 10.5281/zenodo.22746533.

This status is the one declared by the repository’s citation file. Peer review is still to come; until it has taken place, saying so here is the only honest way to put it.

History

• 27 July 2026 — version 2.3 of the text
• 14 September 2026 — Zenodo deposit, DOI assigned
• 21 September 2026 — public opening of the site
• 22 September 2026 — redesign of the site interface

Author identifiers: ORCID 0009-0008-4054-4551 · LinkedIn profile

Around this site

• Use of artificial intelligence — what the assistant does, what it does not do, and to whom your question is passed
• Contribute — the three routes to correct, contest or complete
• Contribution licence
• Invited members
• The diagram as text
• List of figures and tables — with a direct link to each one in the book

How to cite

Cite the book by its DOI, which designates the deposited edition. The data version identifies the extraction these pages derive from; it only serves to reproduce the map.

DOI: 10.5281/zenodo.22746533 · version 2.3 of the text, 27 July 2026

Data version: f7826423b554

This code is the fingerprint of the files from which these pages are derived — the book and the data extracted from it. It changes when their content changes, and only then: two pages bearing the same code were produced from the same sources.

Copy as is:

Yaïche, Nizar (2026). The Humanity Compact. Version 2.3. Zenodo. https://doi.org/10.5281/zenodo.22746533

Citation file of the repository (CITATION.cff) · BibTeX file (.bib) · RIS file (.ris) · OSF mirror deposit

In BibTeX:

@misc{yaiche_humanity_compact,
  author    = {Yaïche, Nizar},
  title     = {The Humanity Compact},
  year      = {2026},
  version   = {2.3},
  doi       = {10.5281/zenodo.22746533},
  publisher = {Zenodo},
  url       = {https://humanitycompact.org}
}

• Full edition, English
• Full edition, French (V2.3)
• Abridged research edition, English
• Annex F — review pack, bounded partial attribution

How the data are derived

Five scripts, in this order. What is entered by hand goes out of date in silence; what is derived corrects itself.

construire_graphe.py

Reads the book’s tables and its structure, and derives the nodes from them. No node is entered by hand; the identifier is derived from the title, never from a counter, so that a URL anchor survives a regeneration.

enrichir_aretes.py

Second pass: resolves the section cross-references in Table 5 and attaches the references to the sections that cite them, requiring a word boundary, a minimum length and a matching year.

extraire_figures.py

Descends to the DOCX XML to attach each image to its caption — python-docx does not link the two — and extracts from it the eleven figures of the book.

calculer_layout.py

Computes the positions for the five views. The rendering therefore does not depend on a force simulation run in the browser.

generer_site.py

Composes these pages and reads the book’s text directly from the DOCX. No displayed figure is written into the code: all are counted from the data at generation time.

What the graph’s links are worth

Of 1655 links, 122 are declared by the book — the author wrote that structure themselves — and 1533 are inferred by the scripts. The diagram draws the former as solid lines and the latter as dotted lines. Each link carries its mode of derivation in the data.

Accessibility statement

What is maintained

• The text of the book is published in HTML, section by section: it can be selected, enlarged and indexed, and read without downloading a five-megabyte PDF.
• Every paragraph carries a stable identifier derived from the section number, never from a global counter: a deep link survives the insertion of a chapter.
• The diagram is rendered in SVG: every node exists in the DOM and in the accessibility tree. A Canvas rendering would have excluded them all from both.
• A complete text equivalent — the table of nodes and their neighbours — is published on a dedicated page, reachable from the diagram and from the footer of every page.
• Reading pages are declared as English, the actual language of their content; editorial pages are in French, and any English title there carries an explicit language attribute.
• A skip link opens every page; the navigation, the main content and the footer are semantic landmarks.
• No external resource is loaded: no remote font, no third-party script, no request to another domain. The site works offline once served.
• The pages are readable and navigable without JavaScript: the view selector, the filters and the search are real links and real forms, and search without JavaScript leads to the text equivalent rather than to a dead page.

What is not yet maintained

• No external accessibility audit has been carried out. What follows is a self-assessment, not a statement of conformity verified by a third party.
• A graph of several hundred nodes remains laborious to navigate with a screen reader, even when properly marked up: the text equivalent is the recommended route, which amounts to admitting that the diagram is not equally usable by everyone.
• The book’s footnotes are not carried over into the reading pages: python-docx does not expose them, and reconstituting them by hand would break the rule that “nothing is keyed in”. They remain in the PDF editions.
• The book’s tables are rendered flat: a merged cell is published once only, without `colspan`, which impoverishes the structure of a few complex tables in the annexes.
• Zooming and panning the diagram have no documented gesture equivalent on touch screens.
• The palette’s contrast ratios target 3:1 as a minimum on both themes, but the measurement belongs to the style sheet and is not checked by this generator.
• No testing with users with disabilities has been carried out.

An accessibility difficulty encountered on this site is a defect of the site, not of its visitor: it is to be reported to the author.