Based on the CX-VE integration discussions one important technical change we agreed is moving CX's content adaptation to cxserver. Currently we do this at browser whenever translator click on a section placeholder in translation view.
The current processing is:
- Source article fetched and shown in first column of Special:CX
- Placeholders shown in second column
- Translator clicks on a placeholder
- cxserver contacted for machine translation of that section.
- At Special:CX, once machine translation recieved, link targets are adapted to target language. References, images, templates, gallery, poem, math.. all these are also adapted to target language
- Translator edit this to improve (using native content editable now, using VE in coming days)
In this step 4 and 5 will be wrapped under a new cxserver api. That will give machine translated & completely adapted section to translator. VE will be then free to work on it as a editor.
Having step 5 was making the whole client side code very complex and demanding more layers in VE integration.