Superba Knowledge — Bridging Regulations and Operations Beta
IT/EN
Download
How we work

Editorial Method

Superba Knowledge is written and maintained by Claudio Leon as an independent initiative, carried out in a personal capacity. It is not a company: no legal entity is involved.

How the pages are built, where the references come from, what happens when something is wrong, and how anyone can check it without taking our word for it.

Independence

No affiliation with the IMO, with flag Administrations, with classification societies, with the regional MoUs, with owners, managers or suppliers. Nobody funds the platform, nobody reviews its content before publication, there is no advertising and there is nothing for sale. Original content is published under CC BY-SA 4.0: anyone may reuse it, including to argue against it, provided the source is cited and the result is redistributed on the same terms.

The twelve layers of a page

All 204 topic pages per language follow the same structure. This is not an editorial affectation: it is what makes it possible to check automatically that no layer is missing, and to put two pages on the same footing when comparing them.

  1. Operational explanation — what it is, addressed to whoever has to do it, not to whoever studies it.
  2. Regulatory reference — the precise provision, with a link to the source.
  3. Scope of application — which ships, from when, with which thresholds.
  4. Procedure — the steps in the order in which they are done.
  5. Practical example — a concrete case, with the figures.
  6. What typically goes wrong — how the procedure fails in practice.
  7. Common mistakes — cross-referenced to the Mistake Library.
  8. What the PSCO checks — what an inspector actually looks at on this point.
  9. Operational tips — what you only learn by doing it.
  10. Preparation checklist — verifiable line by line.
  11. FAQ — the questions that actually come in.
  12. Related topics — the links to neighbouring pages.

Above the twelve layers sit two fixed elements: an opening lede, which has to say in two lines what the topic is and is the same text used in the meta description and in the twin page in the other language, and a row of tags. A check verifies that lede and meta description do not diverge, and that figures quoted in the Italian lede match those in the English one and appear in the body of the page.

Where the references come from

Every link to a source carries a declared type in the page source, because not all sources are the same thing:

Two rules apply across the whole corpus. Where a page prescribes, it must cite the precise provision: the Regulation, the paragraph, the circular with its number — not “SOLAS” on its own. And where no binding instrument exists, the page says so, rather than making a recommendation, an industry practice or a commercial requirement look like an obligation. Pages that prescribed without citing a precise provision numbered one hundred and eight when the check was written; today there are none, and the check keeps it that way.

What happens when something is wrong

It gets corrected, and written down. The corrections log is public and records every round with what was changed and why. The log also states when an external finding was unfounded, with the evidence from the primary source, and when the error was in the correction work itself — those are the most useful entries to read, and the reason they are not removed.

How to verify the page you are reading

At the foot of every page there is a stamp with the date of the last substantive change and a twelve-character fingerprint. The fingerprint is the sha256 of the page content, with the stamp itself removed and whitespace normalised: it establishes that the page you are reading is exactly the one the log describes, without having to trust the date. The stamp is present on 426 pages, and where no round has touched a page it says so, rather than claiming a review that never happened.

For the same purpose, version.txt publishes the fingerprint of the entire HTML tree and of the derived files. Every one of those figures can be recomputed from the site zip: the method is written in the file itself.

The eight automatic checks

Before every publication the whole corpus goes through eight checks, and none of them is a statement of intent: either they pass on all 402 pages, or the round does not close.

  1. Well-formed source: nesting, malformed entities, broken links.
  2. meta description diverging from the lede.
  3. Figures that diverge between an Italian page and its English twin, or that appear in the lede and not in the body.
  4. Pages that prescribe without citing a precise provision.
  5. Expected instruments missing from the module that should carry them.
  6. Maritime Wiki entries not marked as editorial.
  7. Missing accents and recurring typos.
  8. Concentration and coverage of the sources cited.

The recognisers used by the fourth check are themselves under test: 41 cases, of which eleven must fail. A check that never errs on a case built to make it err is not checking anything, and this rule came out of a check that read Italian better than English without anyone noticing.

What this method does not guarantee

How to report an error

If you find a wrong reference, a superseded threshold or a statement that does not hold against the source, write to info@superbaknowledge.org naming the page and, where possible, the provision that contradicts it. Well-founded corrections enter the next round and are written into the log with a statement of what was wrong. The log explains the procedure in full.