But your lsof is equivalent to about 3 lines of rc script, done almost entirely by programs written in C, and is much clearer. I doubt that there's any noteworthy difference in speed, or that it would matter much if there were, so I just can't imagine why you'd want to write it in C.
either i missed a smiley or you missed my point. lsof is not intended for the Plan 9 distribution. if you want to argue, however, here are a few points: your script as you've supplied it is incorrect. adding the required functionality will make it more complex: - results are sorted alphabetically, not numerically by process id (in other incarnations they'll need to be sorted by node, then by process id) - /proc/trace is not a directory - race: at least one process has exited by the time the script tries to cat its fd file, resulting in an error as far as speed is concerned, the rc script is two to three times slower on a single user machine, four to six times slower on a busy machine with multiple users. if one is doing this on several machines at the same time it becomes too much (think 'ps' on a 1000 node cluster).
