Here's my problem: during XSLT processing to generate FO instances I need to be able to make a reasonable guess as to the rendered length of a given string so that I can do some simple automatic copy fitting, such as setting the list label width to fit the longest item in a definition list. I know this can't be exact because of variables like kerning and justification, but in the normal case of unjustified, relatively short strings I should be able to get within an Em width of the correct length.
I think I should be able to do this using the font metric support in the current development version of FOP but I'm having a hard time figuring out exactly what bits are involved.
If you like contorted XSLT and abysmal performance, just use the metric XML files directly. The content should be fairly obvious.