Writing ${\dot X^a}^b$ produces a very cryptic "double superscript" error: this has been explained here on this forum: the reason is that {\dot X}^a generates an Acc atom and that, for a reason that only Knuth must know, putting braces around a single Acc atom does not work to turn it into an Ord nucleus like it does for any other math list (see page 291, ¶5, of the TeXbook, quoted in the aforementioned thread).
What I'd like to know, now, is why $\mathord{\dot X^a}^b$ (or, for that matter, ${\mathord{\dot X}^a}^b$) doesn't work any better: unlike simple grouping, \mathord is not supposed to have any weird exception for single Acc atoms, at least if I believe the TeXbook (just two paragraphs below).
It's simple enough to work around ($\mathord{{}\dot X^a}^b$ works, or simply ${{}\dot X^a}^b$), and of course one shouldn't be writing this sort of thing anyway, but what I'd like to know is: does this count as a bug in TeX? Do I win a $327.68 check from Knuth ;-), or is there some fine print somewhere in the TeXbook or elsewhere that explains how \mathord is supposed to work?
\dotaccent produces an Ord atom anyway …? – Qrrbrbirlbel Mar 12 '13 at 17:10\dotreally seems to produce an Acc atom (why do you think it would be Ord?). – Gro-Tsen Mar 12 '13 at 17:23\dotis an “Acc” atom.\dotmakes\dot Xan Ord atom. TeXbook: “Other kinds of atoms, which arise from commands like […]\mathaccent[…], are all treated as type Ord […]” (p. 170). – Qrrbrbirlbel Mar 12 '13 at 17:28<math field>is. – Hendrik Vogt Mar 12 '13 at 18:16\mathord{<math>}is exactly the same as{<math>}. See also module 1186 intex.webwhere the "brace removing" around an Acc atom is described. – egreg Mar 12 '13 at 19:15