Like any online system worth its salt, especially any that are encapsulating data with XML, our own workflow is strictly Unicode compliant. We require this of our out-of-house data converters, with generally pleasing results.
However, the interpretation of unusual characters by some of the employees of one firm is, well, unusual. CRB has told me of an author who, when trying to cite the Cyrillic-titled journal
Химия Древесины , or KH-EE-M-EE-YA D-R-E-V-E-S-EE-N-YI (Chemistry of Wood)
found the nearest Latin characters and came up with KNMNR DPEBECNHNN (for an alternative Cyrillic spelling) which he then blithely put down as a journal title. What’s most miserable about this is that the author thought that Kunumunur Duhpebecnhhnnn was a perfectly reasonable thing to write.
Our converters, on the other hand, combine naivety with diligence, and will see a Greek alpha (α) and render it as the much rarer IPA character for the open back vowel (ɑ, a stylized “a” without loop or finial, which owing to lack of large-scale support may appear as an empty box in most browsers). Or, more recently, they chanced upon a c-cedilla in “française” (a giveaway if ever there was one) and decided it was the obscure African “c with hook” character (ƈ—and if you can see that character, then you’re driving too close).