I think that we might want to look ahead to the time of having freely available DVI readers in which we can search for symbols inside math zones. How might that work? Point the "mouse" (or the light pencil or ...) at the symbol. If there is an underlying and "identified" symbol, a name for that symbol in ordinary word-characters(*) (in the sense of international character sets) is disclosed by the DVI reader to the user in some way. In order for that string to be disclosed, the author would need to have provided it; for the most part it is not there now except possibly in someone's dvi special. When I pointed to the Connolly note several days ago, it was because Connolly there takes the attitude that a character is any "atom" of information. Mathematicians clearly need to be able to create their own characters in this sense. So, for example, in gellmu one might have, along the lines suggested by Roozbeh, \mathsym{\imag}{\mbox{i}} \mathsym{\ii}{i} where \imag and \ii can only be used inside math (but since gellmu is really SGML one relies on the validator rather than on \ensuremath), and the word-character strings "imag" and "ii" become the names of searchable symbols (if that is what the author wants). (Of course, we all know that the name "\i" is already taken.) Note, however, these two examples might get merged in some presentation formats such as "text/chalk-talk". The optional third argument of \mathsym is for semantic information that could be used locally on an author's platform to make things come out right in MathML. -- Bill * The notion of word-character in the context of international character sets exists in GNU Emacs, version 20+, for use with regular expressions, i.e., \w matches any word character and \W matches any non-word-character. (There still may be the opportunity at this stage for quibbling about the exact nature of the concept.)