Generating well-tagged PDF documents

The talk is devoted to typical problems and possible solutions of generating PDF/UA compliant (or at least well-tagged) PDF documents from authoring applications such as rich layout WYSIWYG text editors, web browsers, and others.
We touch on some ambiguous PDF/UA requirements and technical challenges one faces when converting declarative html-like documents to Tagged PDF. This includes challenges of page-based media versus continuous content flow, and (non)interactive form puzzles. We pay special attention to the areas where PDF 2.0 provides significant improvements, as well as point out where PDF 2.0 does not seem to be sufficient for particular real-world examples.