Thanks! Actually, so far my quick hack is holding up pretty well, the program seems to be running without errors inside its chroot after a few days, so maybe my first take managed to track down all the properties that were needed.

I thought that maybe the -C0 switch would turn off Unicode altogether, which would be OK for this particular script, but for some reason it still doesn't. In the debugger I tracked it down to some data from the network being encoded as ASCII with Encode, which for some reason set the utf8 flag.

I also found a module Unicode::Tussle that had some ability to load up all entities, but really it doesn't seem any less kludgey than the ideas you provide above.

Thanks again!

-----ScottG.