[Post 14 of the DITA Loc Wire series] End of June, we published the sixth article in our series: Six reasons for preferring raw DITA to XLIFF for localization. We concluded that the raw DITA format was better than XLIFF to transfer content between the CCMS (Component Content Management System) and the TMS (Translation Management System).

This article happened to be controversial, while for us the choice was obvious. We contacted one of the technical writing experts who disagreed to try and work out our differences. Both of us realized that the disagreement was not about the file format. It came from a difference in architecture models and processes.

Two architecture models prevail

CCMS TMS architecture DITA localizationModel 1 – CCMS-centric

Its main principles are:

  • The translation review or validation takes place on the published output of the CCMS
  • The target language content is updated in the CCMS which holds multilingual reference content
  • Each single XML element has a unique ID
  • Only the text from updated elements is sent for translation, together with contextual information
  • Translation can be performed using any tool, such as a TMS or a simple Translation workbench, hence the choice of the XLIFF format.

Model 2 – TMS-centric

Its main principles are:

  • The translation review or validation is performed in the translation environment. Efficient review systems use bilingual HTML republishing using DITA OT, thus requiring the DITA content
  • The multilingual content is updated in the Translation Memory hosted by the TMS, which holds multilingual reference content
  • The content repository, which can be an advanced CCMS, can send all raw DITA content without any filtering. The TMS will regenerate the last revised translations.

Both solutions have their advantages

CCMS-centric

  • Facilitates working with any translation vendor, even if they do not know DITA
  • When a mistake is detected in the published content, a simple correction of the corresponding topic and republishing solves the issue.

TMS-Centric

  • The translator learns from the corrections of the reviewer
  • The translation memory is up-to-date with only approved translations, which means that the translator does not reiterate the same mistake
  • The linguistic review is done in a bilingual format that prevents language-specific versions from diverging from the source version
  • Style guide and terminology evolutions are applied based on the Translation Memory maintenance
  • The localization process does not depend on the CCMS ID system, and it ensures that migration from a CCMS to another will work smoothly.

 Which model works best for you?

 CCMS-centric

We recommend opting for this model if :

  • You do not have local Subject Matter Experts
  • You can afford to publish approximate content
  • You can rely on your users to identify and suggest corrections to the content so that its quality improves rapidly
  • You set up a crowdsourcing process to let your users contribute to your content
  • You want to work with many translation vendors.

TMS-Centric

We recommend opting for this model if :

  • You want to publish only high quality approved content
  • You prefer a high-value add partnership with your translation vendor
  •  You want to improve the quality of the published content continuously.

Obviously, as DITA localization experts, we have embraced the second option and strongly advocate for the TMS-Centric model, but a CCMS-centric one could be a good solution in some environments.