> Custom character semantics are very costly in terms of support, and > significantly limit users with respect to the selection of applications > that can be used.
Not to mention that locale and/or "code page" based solutions are awful in that they either bind the language to a document (a single document must be expected to written solely using that one locale/page) or having to use some higher level markup to separate different locales/pages. In this particular case trying to use "@" as a letter character would lead into all kinds of pain if the document(s) were then further processed with software that knows nothing of this new locale/code page. I'm painfully reminded of the various "national ASCIIs" of the 1980's that "reused" various punctuation of ASCII to be their national letters.
