My guess (due to experience with my own network code) is that the module you are using is directly doing send() on the socket object in this case. send() (almost) always expects character codes below 256, otherwise you'll get a "wide character" error. Depending on the exact implementations, those characters may be dropped or an error might be raised.

Sidenote: Proper Unicode handling depends heavily on the age of the module, the experience of the author and also partially if the author is a native english speaker living in a country that is using mostly english to communicate. Basically, if the author is using Umlauts or Emoji or Chinese/Japanese/Korean script on a daily basis, the chances Unicode is handled properly by their code is much higher.

An easy way to really stress-test modules for their Unicode handling is Acme::Umlautify.