3  The Header

Three records that say what the file is, who made it, and which schema it claims to follow — and the one record in the data section that repeats the claim.

The header is short, its shape is fixed, and it is the only part of the file whose meaning does not depend on the schema. It is also the part you will read most often by eye, because it answers the two questions that come up before any other: what wrote this? and what does it claim to be?

HEADER;
FILE_DESCRIPTION(('Open CASCADE Model'),'2;1');
FILE_NAME('block.step','2026-09-17T16:14:23',('J. Chultarsky'),('Physna'),
  'Open CASCADE STEP processor 7.9','Open CASCADE 7.9','Unknown');
FILE_SCHEMA(('AUTOMOTIVE_DESIGN { 1 0 10303 214 1 1 1 1 }'));
ENDSEC;

Notice first what is missing: no #n. Header records have no instance names, because nothing refers to them and they refer to nothing. They are not part of the graph. There are exactly three mandatory ones, they must appear in this order, and each has a fixed set of parameters.

3.1 FILE_DESCRIPTION

FILE_DESCRIPTION(('Open CASCADE Model'),'2;1');

Two parameters. The first is a list of free-text strings describing the file’s contents; exporters put their own name there, or a product name, or leave it as (''). Nothing reads it.

The second is the implementation level, and it matters, though not in the way the digits suggest. The first number is the registered version of Part 21 whose features the file uses; the second is the syntactical conformance class. It is not the edition of the document the writer worked from. '2;1' says the file stays inside what edition 1 defined: one data section, no optional header records, no short names for enumeration values. That is what virtually every file in existence declares, including files written long after edition 2 appeared in 2002. A file that uses edition 2’s multiple data sections or optional header records declares '3;1', and a file that uses anything edition 3 added declares '4;1', '4;2' or '4;3' according to which of the three conformance classes it needs. Chapter 18 covers what changed; until then, '2;1'.

3.2 FILE_NAME

FILE_NAME('block.step',                             /* name */
          '2026-09-17T16:14:23',                    /* time_stamp */
          ('J. Chultarsky'),                        /* author */
          ('Physna'),                               /* organization */
          'Open CASCADE STEP processor 7.9',        /* preprocessor_version */
          'Open CASCADE 7.9',                       /* originating_system */
          'Unknown');                               /* authorization */

Seven parameters, and two of them are the most useful strings in the file.

The name is whatever the writer chose — often the original filename, often with a full Windows path, sometimes empty. The time stamp is ISO 8601, local time with no zone unless the writer adds one. Author and organization are lists of strings and are usually ('') and (''). Authorization is a free string that almost nobody fills.

Preprocessor version names the software component that wrote the STEP file — the translator library. Originating system names the application it was running inside. They are distinct because most CAD systems do not write STEP themselves; they license a translator. When the two strings are the same product, as here, the vendor wrote their own. When they differ, you learn something:

Preprocessor Originating system What it tells you
Open CASCADE STEP processor 7.9 Open CASCADE 7.9 OCCT, or a tool built on it (FreeCAD, CadQuery, many others)
ST-DEVELOPER v20 NX 2312 Siemens NX, writing through STEP Tools’ library
SwSTEP 2.0 SolidWorks 2024 SolidWorks’ own translator
CATIA V5 STEP AP214 CATIA Version 5-6 ... Dassault, own translator
Pro/ENGINEER by Parametric ... Creo Parametric 10.0 PTC, own translator

These are the strings as they characteristically appear; version numbers change every release and the exact wording drifts. The point is that once you know a file came from NX via ST-Developer, or from SolidWorks, you know a great deal about which of the structures in Part I and Part III it will contain and how it will spell them. Every experienced STEP user reads these two fields first.

3.3 FILE_SCHEMA

FILE_SCHEMA(('AUTOMOTIVE_DESIGN { 1 0 10303 214 1 1 1 1 }'));

One parameter, a list of schema names, almost always with one element. This is the line that says which application protocol the file follows. Every record in the DATA section must be an entity type from this schema.

The string has two parts. The name before the brace is the schema’s EXPRESS identifier, upper-cased. The braces hold an object identifier — a sequence of integers registered with ISO that pins the exact schema and edition. The ones you will see:

Schema name Protocol
CONFIG_CONTROL_DESIGN AP203, first edition
AP203_CONFIGURATION_CONTROLLED_3D_DESIGN_OF_MECHANICAL_PARTS_AND_ASSEMBLIES_MIM_LF AP203, second edition
AUTOMOTIVE_DESIGN AP214
AP242_MANAGED_MODEL_BASED_3D_ENGINEERING_MIM_LF AP242

The _MIM_LF suffix on the newer names means module interpreted model, long form — the AP’s schema flattened into a single file rather than assembled from modules, which is the form tools actually load. The object identifier’s digits differ between AP242 editions, which is how a reader can tell an edition 1 file from an edition 3 file when the name is the same.

The same block, exported by the same OCCT build with the schema set to AP242, has this header and a slightly different pair of opening records:

FILE_SCHEMA(
  ('AP242_MANAGED_MODEL_BASED_3D_ENGINEERING_MIM_LF {1 0 10303 442 1 1 4 }'));
ENDSEC;
DATA;
#1 = APPLICATION_PROTOCOL_DEFINITION('international standard',
  'ap242_managed_model_based_3d_engineering',2013,#2);
#2 = APPLICATION_CONTEXT('Managed model based 3d engineering');

Every other data record is identical, apart from #7, whose identifier and name were left as the exporter wrote them in this file. For a plain solid there is no difference between an AP214 and an AP242 file beyond the header and the three values in #1 and #2 — the schema name, the year and the context text — which is why a reader that handles one handles the other with little extra work. The differences begin with colours, PMI and tessellation.

WarningIn the wild: When the schema lies

FILE_SCHEMA is a claim, and exporters break it in two ways. Some declare CONFIG_CONTROL_DESIGN (AP203, which has no colour entities) and then write AP214 colour records anyway, because that is what the user asked for. Others declare AP214 and write entities that exist only in AP242. Strict readers reject these files; practical readers treat the header as advisory and accept any entity type they recognise from any of the three protocols. There is no rule in the standard that permits this, and it is what every widely used reader does.

3.4 The optional header records

Edition 2 defines three more header entities that may follow FILE_SCHEMA:

FILE_POPULATION('AUTOMOTIVE_DESIGN','SECTION_BOUNDARY',$);
SECTION_LANGUAGE($,'eng');
SECTION_CONTEXT($,('...'));

FILE_POPULATION says which instances, gathered from which data sections and by which of three named methods, are to be checked against a schema; SECTION_LANGUAGE declares the natural language of the strings, as a three-letter ISO 639-2 code; SECTION_CONTEXT is a free-form context list. They are rare enough that many readers do not implement them and a few reject files that contain them. Edition 3 adds more. The standard allows user-defined header records as well, provided the keyword begins with an exclamation mark, as in !A_SPECIAL_ENTITY('ABC',123);. Anything else in the header section is an error — and some exporters put things there regardless.

3.5 The record that repeats the claim

Strictly, the header is finished at ENDSEC. But the first two records of the data section belong with it in spirit, because they exist to say the same thing the header said:

#1 = APPLICATION_PROTOCOL_DEFINITION('international standard',
  'automotive_design',2000,#2);
#2 = APPLICATION_CONTEXT(
  'core data for automotive mechanical design processes');

APPLICATION_CONTEXT is the root from which the product-identity records hang: every PRODUCT_CONTEXT and PRODUCT_DEFINITION_CONTEXT in the file points back at it, and it says, in prose, what domain the data is in. APPLICATION_PROTOCOL_DEFINITION attaches to it and records the AP by status ('international standard'), name ('automotive_design') and year (2000).

Why repeat what FILE_SCHEMA already said? Because the header belongs to the file and these two records belong to the data. A file might in principle carry data from several contexts; and when data is moved out of Part 21 into a database or another encoding, the header does not travel with it but the records do. In practice there is one application context per file and it agrees with the header. When it does not, believe the header, and note that the exporter was careless.

Each AP’s schema has a global rule fixing the schema-name string in #1, and it differs by edition. AP203 edition 2’s rule requires 'ap203_configuration_controlled_3d_design_of_mechanical_parts_and_assemblies' — the long name, but without the _mim_lf that its schema name carries. Edition 1 files carry 'config_control_design' instead. AP214 files write 'automotive_design', as above, which is the lower-cased schema name. AP242 requires 'ap242_managed_model_based_3d_engineering_mim_lf' — and the export shown earlier writes it without the _mim_lf, which is a rule violation of exactly the kind Chapter 1 warned about. The year and the APPLICATION_CONTEXT text are convention: nothing in the schema checks either. They are the second place, after the header, where you can identify the protocol — and a place where files that have been converted between protocols often still show their origin.

ImportantIn brief

Three header records, no instance names, fixed order. FILE_DESCRIPTION gives the implementation level: the Part 21 version whose features the file uses and its syntactical conformance class, not the edition the writer worked from. FILE_NAME names the translator and the CAD system, which predict the file’s quirks. FILE_SCHEMA names the AP, and is a claim rather than a guarantee. APPLICATION_CONTEXT and APPLICATION_PROTOCOL_DEFINITION, the first two data records, repeat that claim from inside the data.