As Kristen mentioned, often dita is processed before being presented in a translation workbench. Whether that processing takes place in the CMS, the desktop or in the TMS import is not of importance. And whether the TMS imports dita or XLiff is a tool and processing issue downstream from DITA. The primary issue for DITA is whether authors can organize their terminology in the source language to take advantage of single sourcing, indexing and glossary mechanisms. The secondary concern is whether the necessary structure exists in DITA for the fully aware toolchain to leverage source terminology structuring and to further support DITA type terminology mechanisms in the TMS outputted target language - DITA content which comes back from the TMS. What mechanisms are required in the TMS to support this – as I think Michael mentioned. DITA considerations in the source content for underperforming tools in the toolchain – are just that, considerations, not best practices or requirements. I really think the ball is in the court of the TMS to address this translation issue and then work backwards to DITA as required.