As part of the edits made in the move from WD (http://www.w3.org/TR/2010/WD-CSS2-20101207/visuren.html#z-index) to PR (http://www.w3.org/TR/2011/PR-CSS2-20110412/visuren.html#z-index), an error was introduced in 9.9.1.

The following WD text:

  # Positioned elements with 'z-index: auto' (in layer 6), floats
  # (layer 4), inline blocks (layer 5), and inline tables (layer 5),
  # are painted as if those elements generated new stacking contexts,
  # except that their positioned descendants and any child stacking
  # contexts take part in the current stacking context.

was changed in the PR to:

  # <ins>Within each stacking context,</ins> positioned elements with
  # stack level 0 (in layer 6), non-positioned floats (layer 4),
  # inline blocks (layer 5), and inline tables (layer 5), are painted
  # as if those elements <ins>themselves</ins> generated new stacking
  # contexts, except that their positioned descendants and any
  # <ins>would-be</ins> child stacking contexts take part in the
  # current stacking context.

Note that, apart from the three indicated edits that were discussed on the mailing list [http://lists.w3.org/Archives/Public/www-style/2011Mar/0108.html], there was also a change from "positioned elements with 'z-index: auto'" to "positioned elements with stack level 0".

This change was not discussed, to my knowledge.  Unfortunately it is incorrect, since positioned elements with stack level 0 include positioned elements with 'z-index: 0', and these latter elements /don't/ have the described "almost stacking context" behaviour.  (Rather, they have full stacking context behaviour.)