Execution speed is critical to the application I'm writing. Any process that relies upon navigating the file system's directory tree will probably be too slow.

It might be, but you won't know until you test it. I suggest you try the easy way - code up a solution that uses readdir() and stat(). If that isn't fast enough then start working on the hard way.

I had also looked into using C instead of Perl. Every C snippet I found seemed to use a different set of header files. In no case did I have all of them present upon my system. I'd prefer to only write the thing once rather than have to port it for different Unix flavors.

I think you'll have to pick your poison: slow and portable or fast and non-portable. As far as I know there's no Unix standard for the binary structure of an inode. If you want to get the best possible speed you'll need to drop to C and access the inode directly. If you do that you can't expect it to work without changes on different OSes.

-sam