Spark saves the log file as UTF-8. Using the Internet Explorer one can read the file.
If Spark reads the file to display the last 15 lines it seems to use another encoding, so it displays bulls**t. While my tests with very short files (4 messages) Spark also saves the file with the wrong-read characters. So the chatlog is very soon a great candidate for /dev/null if a lot of UTF-8 characters are used (chinese, russian, …).
For our US ASCII only chatters who want to reproduce this on Windows: Start, Programs, Accessories, System Tools, Character Map and then select as character set not “DOS: United States”/“Windows: Western” but “Unicode”.