DIN, ISO, NF, EN: what each prefix means, and what the engine isolates

Who publishes what

ISO is the International Organization for Standardization; EN is a European standard, published by CEN — the European Committee for Standardization, an association bringing together the national standards bodies of 34 European countries (cencenelec.eu, “About CEN”, read on 20 October 2026). NF is a French standard, approved by AFNOR. DIN is the German body, the Deutsches Institut für Normung.

The point that matters for a catalogue is not the body: it is that the prefixes stack. AFNOR describes the mechanism as Russian nesting dolls: an international standard adopted in Europe and then in France has its title extended with the matching prefixes, in ascending order of geographic scope — NF ISO, then NF EN ISO. And when a standard carries only the NF prefix, it applies in France alone (AFNOR, FAQ on the world of standards, read on 20 October 2026).

The same source gives the order of magnitude, and it explains why a French technical catalogue is full of ISO and EN: 90% of voluntary standards are of European or international origin, and only 1% of standards are mandatory.

One asymmetry decides the rest. A European standard must be adopted as a national standard by member countries; adopting an international standard, by contrast, is voluntary. DIN puts it most plainly, writing about its own mirror committees (din.de, “What is a DIN Standard?”, read on 20 October 2026). Hence the composite designations a buyer copies straight off a drawing: “NF EN ISO 4014”, “DIN EN ISO 4762”. Three prefixes, one number, one object.

One source we could not read, and it is on the record: on 20 October 2026, iso.org returned an HTTP 403 refusal on its own robots.txt. Our rule for reading a third-party site requires reading that file first and honouring it; a refusal on the file that states what is permitted permits nothing. No page of iso.org was read, and no claim in this article comes from there.

What the engine can isolate

A standard inside a product reference is only useful to search if it becomes an attribute of its own: otherwise “DIN 933” is just a run of characters, and the three ways of writing it are three different products. Our hardware pack carries three standard rules, and only one of them covers French standards.

In-house measurement. Labels run through the annotation cascade of the outillage pack on 20 October 2026, engine at commit d25011d, each label run twice — Python cascade and production Rust core — with the same result throughout. Only standard annotations are reported here; the labels produce others, for diameter, length and head type.

The first three rows are the service being rendered: “DIN 933”, “DIN933” and “DIN-933” return the same annotation, so the same attribute, so the same product findable through all three spellings. Fault tolerance is not what does this work, the annotation is. You can see it in the result explanation: on the query “din933”, the screw labelled “DIN 933” comes back with NORME_DIN933 as its only justification. No term matched.

The part-suffix row states a decision: “NF E25-030-1” returns NORME_NFE25_030, the suffix is dropped. Two parts of one standard are the same standard to whoever is searching.

EN is not a standard to the engine, and it shows

None of our packs carries a rule for the European prefix — verified against the repository's complete inventory of standard rules: three in the hardware pack, one in the electronics pack, none for EN. Two consequences are measurable:

On “EN ISO 4014”, the engine returns NORME_ISO4014. It reads the ISO part and ignores the prefix. The result is right, by the ISO rule alone — and it is right because the composite designation contains the ISO number, not because the engine understood the stacking. On a purely European standard there is nothing left to read: “EN 10204” — the standard for inspection documents on a metallic product — returns no standard annotation.

V4  Vis tete cylindrique ISO 4762 M6x30 inox A2
V6  Vis tete hexagonale EN ISO 4014 M8x20

query “iso 4014”     ->  V6 22.0   V4 4.5
query “en iso 4014”  ->  V6 28.0   V4 4.5

Both queries find the right product. The second finds it better, for a reason that has nothing to do with standardisation: “en” is one more term matching the label. The prefix counts as text, not as a scope.

NF: one series, and a deliberate exclusion

The French rule reads one format only: the prefix, the letter E, two digits, a hyphen, three digits. That is AFNOR's E series, the one covering screws and bolts. All three trade spellings pass, and it is not the pattern that absorbs the spacing variants: it is the tokeniser, which keeps the hyphen inside the token and only splits on spaces, doubled by the two views of the text — spaced and compacted — that the cascade applies.

The letter is bounded to E, and that is a decision written into the pack: “NF C 15-100”, the standard for electrical installations, has exactly the same format and has no business in a fastener catalogue. Measured: that reference returns nothing in the hardware pack. But the pack's comment went on to say the electrical pack would read it if someone wrote it there — and that is false today. Run through the electrical pack, the string “Coffret etanche IP65 NF C 15-100” returns FAM_COFFRET · IP_65, and nothing of the standard. The electrical pack carries no standard rule at all. The discrepancy is in the engine, not on this page, and it has been reported.

Last row of the table: “NFX11443” returns nothing. It is not a standard, it is a reference from a supplier whose name begins with NF. It has no hyphen, and it is the hyphen the rule requires that rules it out.

The same rating, two packs, two names

One case is worth knowing before you build a facet. The IP ingress-protection rating is read by two packs, under two different annotations:

The same string, “IP65”, becomes IP_65 on one side and NORME_IP65 on the other. Both are right within their vertical, and a hybrid catalogue loading both will carry both on the same product. This is not a defect — it is the kind of thing to know before naming a facet, because the annotation's name is what a filter is written against. The exact shape of every annotation our packs can produce is published by the engine and documented in our annotation cascade documentation.

Two limits we accept

The engine does not link two standards together. That a DIN reference and an ISO reference name the same part is a correspondence between two standards catalogues, not a shape readable in the text. No pattern produces it, and no engine release will: it is not the same kind of object.

The standard rules live in one pack. The three rules measured here are in the hardware pack. A plumbing or electrical catalogue that cites standards in its labels will not see them isolated, unless it also loads that pack — a catalogue accepts several — or a custom rule is written. A custom rule teaches the engine a convention it does not yet know; it does not link two bodies together.

What to do about it

If your labels already carry their standards as prefix plus number, they become an attribute with no configuration, and all three spellings of each one converge. If your references are composite designations, the ISO number is what gets read, and the rest counts as text. And if your catalogue is fasteners, everything else the hardware pack reads — diameter, length, head, material, grade — is what does the work around the standard: our Hardware & tooling page details it attribute by attribute.