From http://www.unicode.org/reports/tr50/ (still in draft at the time of opening this bug) "When text is presented in vertical lines, there are various conventions for the orientation of the characters with respect to the line. In many parts of the world, most characters are upright. In East Asia, Kanji and Kana characters are upright, Latin letters of acronyms are upright, while words and sentences in the Latin script are typically sideways. This report describes two Unicode character properties which can be used to determine a default orientation of characters in those two scenarios." We need to support the relevant character properties and handle glyph orientation in vertical text flow.

To give more specifics to this, when text orientation is not explicitly set to something other than the default, user agents need to break up a string into vertical and horizontal runs, based on the orientation property of the codepoints defined in UTR50. Ex: ＮＡＳＡ（National Aeronautics and Space Administration）は、１９５８年７月２９日正式に発足しました。 When displayed vertically, this will be sliced into: vertical: ＮＡＳＡ（ horizontal: National Aeronautics and Space Administration vertical: ）は、１９５８年７月２９日正式に発足しました。 Note that NASA and ＮＡＳＡ use different codepoints with different default orientation. ASCII letters and numbers are laid out horizontally and rotated 90deg clockwise while kanji/kana and full-width codepoints are laid out vertically. The default orientation can be explicitly overridden with 'text-orientation'.

I should note here that layout code needs to understand how these subspans are broken up because things like selection and editing will behave differently within subspans with different orientation.

AFAICT, this can be considered done, via bug 902762 and its dependencies (note particularly bug 1057330 and bug 1057331). Closing as a dupe of bug 1057331, where we actually break up the spans as needed for different glyph orientations.