Isn't it technically possible to use any character encoding in clamspeech? Whether or not UTF8 is supported seems to be an issue for whatever is parsing/presenting the clamspeech. After all, the bytes are there, it's just a matter of encoding them as characters for human consumption.
Yes, that's what my patch does. It just copies the bytes from the clamd client to the server and to into the transaction, no conversion at all (should be tested on Windows as well, I tested it only on Linux and with the command line tools, not the Qt client). You can use ISO-8859-1 if you like, or UTF8, or any other encoding.
But the transaction doesn't store the information which encoding you used, so as a convention it would be good to use only one encoding. UTF8 would be the best, I think, because it is the international standard nowadays and you can encode nearly anything with it.