Le 08/01/2014 23:52, C. Scott Ananian a écrit :
One option here is to get serious about the 'post processing' stage of parsoid. The goal was to have a stripped down HTML-based representation for MW content. But many applications (kwix, pdf rendering, etc) want some additional "sugar" in the HTML. They would like, say, more precise link types, or alt and title tags on images, or redlink information. I believe gwicke's original plan was to have a post processing pipeline where this could be added.
Perhaps some of that postprocessing could optionally take place within the parsoid codebase. If you requested "elaborated html" from parsoid, it could go through and add some extra attributes and rel types. Would that sort of design satisfy your needs, Emmanuel (and allow us to remove the rel attribute in the core representation)?
If this is possible to offer both, like you propose, without generating too much additional work, then this looks to me to be an excellent proposition.
Emmanuel