We propose a general strategy for deriving 7-bit encodings for texts in languages which use an alphabetic non-Roman script, like Arabic, Persian, Sanskrit and many other Indic scripts, and for which there is some transliteration convention using Roman letters with additional diacritical marks. These schemes, which we will call ``meta-transliterations'', are based on using single ASCII letters for representing Roman letters, and digraphs consisting of a suitable punctuation character and an ASCII letter for representing letters with diacritics. A meta-transliteration is required to be uniquely reversible, human readable, and close to the intended transliteration. We present an example of a scheme that has been in use for several years to transliterate texts in Arabic, Persian, Urdu, Sindhi, and Biblical Hebrew.