Only Latin JIDs REGEXP

Note that erlang treats "\" as its own escaping character. So if you need to include the "\" into the regex pattern (say, if you need "-" inside square brackets), then you need to double-escape this in ejabberd.cfg:

(I put the dot here unescaped, as it is not a metacharacter inside square brackets, but no harm if you will escape it, too)

Also note that ejabberd.cfg is utf-8-encoded, and must not include the BOM. This is important if you use internationalized characters in it (like in your initial pattern "[а-яА-Я]"). If your editor includes BOM, or if it saves your file as ACSII, then ejabberd may not work properly. I use Cyrillic characters in my ejabberd.cfg (my users, to the contrary, almost exclusively use Cyrillic characters in their uids), and it works. Still, I don't know if this works with regex, because in utf-8, internationalized characters are multibyte, and this will work only if the string is internally converted to utf-16 and then passed to regex module, or if the regex module knows how to treat multibyte utf-8 sequences. This needs trial. Possibly you may need to use "\x{ABC}" to succeed in this case.