Will an input-oriented (recognition mode) Nymalizer be able to differentiate between the outputs of "actual people" (ha) and actual people as filtered through output-oriented (generation mode) Nymalizers? After all, once we discover that username 'x' posts using LOTS of CAPITALIZED words and phunny SPELLINGS and so forth, it doesn't seem so hard to turn a Nymalizer into a filter which would create (or at least describe to a human creator) text which would fit that pattern .. I'm pleased to see that Perry now supports moving politics off the list; his posts do make me wonder if he's being spoofed, as he cheerfully participated in the econ-flames of a month ago. Still, I wholeheartedly support either him or the spoofer in their call to take the politics somewhere else. -- Greg Broiles greg@goldenbear.com Baked, not fried.