Given that Unicode is pretty serious (at least when sticking to principles) about attempting to keep glyphs as assigned codepoints distinct from visual representation (not just in the “FFS people! The font in our document is just a placeholder, not normative!”; but in things like “‘GREEK CAPITAL LETTER DELTA’ (U+0394)” being treated as wholly distinct from “‘MATHEMATICAL BOLD CAPITAL DELTA’ (U+1D6AB)” because the two have completely different meaning despite the latter being a straight appropriation of the former) I would imagine that “reversed glyph” would be a distinct thing from “switch to right-to-left”(though it would still most properly be a modifier, rather than a whole bunch of mirror-mirror codepoints; since the Unicode way is supposed to be doing things by composition rather than addig scads of additional primitives, where possible): “reversed glyph” would just mean rendering that character backwards without other modifications; while switching to right-to-left means adopting the relevant conventions for left vs. right justification and anything else that makes a right-to-left language distinct from left-to-right with the symbols backwards.
For very short snippets, not long enough to trigger any of the additional rules, having a stash of backwards glyphs would be a hack that makes that case a lot simpler; but it would lead to messiness in other respects (probably most annoyingly having both right-to-left and reversed-glyph-hack would mean that apparently identical strings wouldn’t match at all; copy/paste behavior would also get pretty freaky if you had to start inferring when to automagic the snippet into reverse glyphs and when to keep it as is but surround with the appropriate direction reversals).
Doesn’t make life easier for the renderer; but when your scope is “encoding” there’s an argument to be made against the sort of hacks that would be just fine if you knew that the output was heading straight to the printer, because you haven’t the slightest reason to believe it is; and a result where things like string comparison breaks down would not be good.
