Concurrent localization model
This section describes the concurrent localization model.
This localization model allows multiple revisions of the source to be in translation at the same time. It is more flexible than the sequential model but it can lead to a much larger repository size and increase the risk of duplicate translations.
First localization of an object
Figure 1 shows the process for the first translation of an object in the concurrent localization model. The first translation of a source object in the concurrent localization model happens exactly as with the sequential localization, but the language object created includes the authoring version number in its name. For example, consider a document named abc1386692219292.xml that is at revision 6. When this document is localized, a file named abc1386692219292_00006.xml is created in each target language.
This language object has its own Revision number as well as an Authoring Revision property, which stores the revision number of the source object when it was sent for translation. Its content is still in the original language (for example, English), but its language attribute is set to the target language. The language object has the tb translated state.
When you prepare the localization kit to send the files to the translation team, the language object still has the same Revision number (for example, 1), but its state is now in translation. The files sent to the translation team are extracted from the Content Store and then converted to the XLIFF format.
When translated content is imported into the Content Store, a new revision of the language object is created (e.g., Revision 2), and its state is changed to review so that the translated content can be reviewed internally. The Authoring Revision is not updated so that it is always possible to correlate a language object with its original source. If the translation is updated during the review process, a new revision of the language object is created (for example, 3).
When the translation is approved, its state is changed to Done, and you can now publish the localized map.
Subsequent localizations
When a new revision of a source object is sent to localization (for example, Revision 9), the CCMS does not update the language object, as with the sequential localization model. Instead, it first checks to see if this revision has already been sent for localization (for example, if it was part of a another map already sent for localization). If so, the localization process skips the file.
If this revision has not been sent for localization, it creates a new object in the target language, and its name includes the authoring version number (for example, abc1386692219292_00009.xml). This new localization object has no impact on the previous translation (named abc1386692219292_00006.xml). This is why this model is concurrent: multiple language objects of the same source object exist concurrently.
The new language object has its own revision number at 1 and its status is set to Localization:tb translated, as shown in Figure 2.
Auto-translation is always enabled for the concurrent localization model. If a previous translation exists with a Localization:done status, the CCMS then auto-translates the content using the translated text from the previous version of the language object (that is, the Localization:done version).
If the language object was fully translated during auto-translation, the CCMS creates a new revision of the language object with the auto-translated content and sets its status to Localization:machine translated (or its equivalent in your workflow, depending on your configuration). Otherwise, the next step depends on whether your system is configured to accept partially auto-translated documents.
If partially-translated documents are accepted, then the CCMS creates a new revision of the language object with the partly translated content and sets its status to Localization:tb translated. Otherwise, the partial translation is rejected and the source content is sent in the localization kit.
With the concurrent localization model, there is no dependency between the various authoring revisions that are sent to translation; a new authoring revision can be sent to translation independently of the localization status of the previous authoring revision.