Technical reference note

Read, Write, Import, Export: Why Support Is Directional and Version-Bound

Interpret format-support statements by operation, product release, representation scope, and documented limitations before relying on an import or export.

Formats, Profiles, and Translators 6 min read

Updated 17 Oct 2026

“Supported” is an incomplete verb. A product may read a representation without writing it, open it without preserving its authoring semantics, or write a version that a particular receiver does not accept. Read support is not a reciprocal promise of write support.

A directional export and import route between two products, each bounded by release and scope and followed by a route test.

Define the operation

Use the product’s own terminology, then record what the operation means in the proposed workflow:

Term Practical question
Direct open Can the product open the source representation through its native or direct reader?
Import or read Can the product ingest the representation into its own working model?
Export or write Can the product serialize a result into the named representation?
Reference Can the product use the source in place or through a linked view without creating the same kind of editable model?
Translation What mapping occurs between the source representation and the destination model or output?

The words are not universal API definitions. A support page’s definition and scope take precedence over assumptions based on another product. In every case, identify the direction: source → destination for an import, and destination model → output representation for an export.

Qualifiers that make a support claim useful

Rewrite a bare statement such as “supports STEP” into a record containing:

  • product name, edition, and release;
  • translator or exchange component version, if documented;
  • operation and direction;
  • source and output format identifiers;
  • schema, application protocol, profile, or model view;
  • source-version range or exact release;
  • content scope and stated limitations;
  • settings or dependencies that affect the result; and
  • the downstream operation that still has to work.

Without these qualifiers, the statement may be true for a different release, direction, representation scope, or content subset.

A bounded historical reading exercise: VRED 2026.2

Autodesk’s VRED documentation has a section for VRED 2026.2 and states that it uses ATF 14.23.0. Its tables separate columns named “Import (Read)” and “Export (write).” In that release section, the table lists these examples:

Representation Import (Read) Export (write) What the row does not establish
3MF (.3mf) All versions All versions That a specific 3MF extension or print workflow is preserved
Inventor (.iam, .ipt) Up to Inventor 2026 N/A That a VRED import remains editable in Inventor
IFC (.ifc) IFC2x3, IFC4, and IFC 4.3 N/A Which view, information requirement, or IFC semantics survive
STEP (.step, .stp) AP214, AP203E2, and AP242 N/A That every entity or downstream CAD operation is preserved
USD (.usd, .usda, .usdc, .usdz) All versions All versions That all composed assets resolve in the receiver
FBX (.fbx) All versions 2016, 2018, 2019, and 2020 That the import and export mappings are equivalent

This is evidence about the VRED 2026.2 documentation and its listed scopes. It is not a general compatibility matrix, a product recommendation, or an end-to-end route result. The page contains historical release sections, so do not mix their rows into a current claim.

Separate a toolkit change from support rows: VRED 2027

Autodesk states that VRED 2027 uses the New Data Translator as its only translator for NX, SolidEdge, SolidWorks, and CATIA V4/V5 files; the Legacy Translator and the switch between toolkits were removed. That release note identifies which toolkit serves those file families. It does not give their read/write direction or version bounds.

A separate VRED 2027 release-note section gives these selected file I/O declarations:

Direction Listed representation/version What the declaration does not establish
Read Inventor 2027; Parasolid Binary 38 and Parasolid Text 38; NX 2506 and 2512 Series; Solid Edge 2026; 3D ACIS ASM 232 That every file at those versions is accepted or that its payload is preserved
Write JT 10.7 That a particular receiver accepts the result or that it retains a required data layer

This is a selection from the VRED 2027 File I/O release note, not a complete support matrix. The same release note documents an option to exclude constraints and physics attachments when exporting a scene or selection to VPB or OSB. A documented option is not evidence that a given export used it.

Read consortium matrices as declarations: 3MF

The 3MF Consortium’s matrix separates Core Import from Core Export and lists extensions in their own columns. It says vendors self-report the entries and that the Consortium does not independently verify them. Its “Custom” label means custom or partial support; an empty cell is not a negative support claim, and the page provides no uniform product-release or extension-version column.

For example, the displayed Fusion 360 row lists Core import and export and Materials import, without a Materials export entry. Read this as a vendor-reported matrix declaration without a product release or file test—not as independent verification, proof that an individual material payload survives, or a statement that the blank cell means unsupported.

Version asymmetry is normal

A writer can emit a newer schema or representation variant than the reader understands. A reader can accept several older variants while the writer exposes only selected output versions. Even when both sides name the same family, a release, profile, extension, or implementation setting can change the result.

The buildingSMART release database illustrates why status belongs in the record: IFC 4.3.2.0 is listed as official, while IFC 4.4 and IFC 5 are under development on the accessed page. glTF likewise requires an asset to declare its target version, while a GLB carries a separate container version. These format-level facts do not replace checking product support.

Questions to settle before transfer

Send the receiver a precise question:

Can product A, release X, export representation Y at scope Z, and can product B, release W, import or reference that same scope in direction D for operation O? Which data is unsupported, transformed, or external, and how will we validate it?

Ask for:

  1. the exact documentation page and product release;
  2. the operation and direction, not just the format name;
  3. version, schema, profile, model view, and extension scope;
  4. known content restrictions and fallback behavior;
  5. dependencies such as external resources or linked files; and
  6. a test and acceptance method for the actual downstream task.

If the answer is only “the file opens,” continue with what a translator actually does and the relevant information-layer validation. A support declaration tells you what to test; it does not perform the test.

Key points

  • Import/read and export/write are directional operations.
  • Product release, translator version, representation scope, and content limits are part of support evidence.
  • A support row does not establish semantic preservation or downstream usefulness.
  • Toolkit availability, directional I/O declarations, and self-reported consortium entries are different evidence types.
  • Treat time-sensitive support documentation as a bounded observation that must be rechecked when the route matters.