An XLIFF file is structured into multiple translation units. While a string property is encoded as a single translation unit, a richtext property is split into semantically meaningful parts, comprising for example a paragraph or a list item. Translation units are then grouped, so that units belonging to a single property are readily apparent.

All properties of a single content item are included in a single file section according to the XLIFF standard. A custom attribute cmxliff:target allows the importer to identify the target content item that should receive the translation, as supported by the XLIFF standard. Translation tools must preserve this extension attribute when filling the target content into the XLIFF file.

For details on the standard and extended XLIFF schemas, see the corresponding Javadoc of XliffSchemaConfiguration. Refer to the 'See also' sections for links to the schemas.

The fragment in Example 5.14, “XLIFF fragment” shows the start tag of a <file> element for translating from English to French, indicating the source document 222 and the target document 444.

<file
  xmlns:cmxliff=
     "http://www.coremedia.com/2013/xliff-extensions-1.0"
  original="coremedia:///cap/version/222/1"
  source-language="en"
  datatype="xml"
  target-language="fr"
  cmxliff:target="coremedia:///cap/content/444">

The elements to translate are grouped in <trans-unit> elements, which consists of a <source> element containing the original text and will later contain the translated text inside a <target> element.