Imported content can change without changing Word text
A Word package can keep all stored text, an import anchor, and its relationship
ID fixed while changing alternate content that a client may later import. That
is a package change a text diff cannot see, but a static reviewer can report it
without attempting the import.
Document Change Assurance Benchmark 0.5.0
adds its sixteenth deterministic pair:
import.alternative_format_html_payload_changed. Both sides retain a fixed
w:altChunk r:id anchor, an internal afChunk relationship, every package
member, and every stored w:t sequence. Only word/afchunk1.html changes.
Inspect the payload; do not import it
Microsoft’s Open XML documentation for AltChunk
describes the anchor as a location for alternate content to be imported through
an internal afChunk relationship. DCAB uses a compact inert HTML marker, but
its builder, validator, scorer, and adapter never parse HTML, import content,
launch Word, or claim what a client will render.
The public truth names only alternative_format_import_payload_changed, with
an HTML payload kind and the Word altChunk source. It excludes payload bytes,
content, relationship IDs, relationship paths, and any client-generated
WordprocessingML.
A deterministic internal boundary
The generator emits the standard internal relationship and HTML content-type
override only when a payload is present. An independent verifier checks the
anchor, relationship type, internal mode, payload member, deterministic bytes,
stable package members, unchanged stored text, and exact one-member boundary.
The standard python-docx reader opens all 30 .docx fixtures, and its
lower-level OPC reader opens all 32 packages.
The optional local DocFence 0.27.0
adapter maps public alternative-format import inventory evidence while anchor,
relationship, and payload-part counts remain one. It does not consume payload
content or private signatures.
DCAB 0.5.0 keeps fixture schema version 1 because the public envelopes are
unchanged; it extends the corpus from 15 to 16 cases. It does not claim imported
HTML is safe, processed, renderer-equivalent, or handled by a client.
The source, generated corpus, and verifier are MIT-licensed on
GitHub. Read the
canonical release note
for the full boundary and install command.
Top comments (0)