> Why not develop a system call for obtaining process info? The same
> goes for IO. Wouldn't it be nice to have a STANDARD api for querrying all
> devices that do IO? Imagine all the utilities that could be written to
> take advantage of information like this?
I only have one comment about this -- this is how windows does it.
One thing I like about /proc and /etc/passwd and /dev/* stuff is
that it's frequently very obvious how to use them. With windows, however
(or even some ioctls in /dev), it's difficult to know exactly how it's
supposed to work until you've tried it. Syscalls in Unix are generally
easier to figure out than in Windows (if not just because we unix people
aren't ashamed to use a pointer without trying to hide it behind a typedef),
and they're well enough documented that they can be used, but I don't want
to see an excessive number of syscalls. ioctl's are almost never decently
documented (at least not in an easy to find way), and their arguments are
frequently confusing.
Hrm. that wasn't very clear. :) I just think that if you provided
ioctl's and left the text support in there, no question about it, I'd stick
with the text in my applications unless a *useful* procps lib were made that
would use the ioctl's. In which case, switching would be easy as pie.
Greg Alexander
http://www.cia-g.com/~sietch/