> >Hmm. I assume the main reason is the extra bookkeeping overhead you
> >describe.
> Ahh. So, what if the kill -9 has the corresponding effect of
> zombing the process?
That is more or less what happens. If ressources get needed, the blocked
process gets swapped out. It isn't zombied literally, but effectively.
> Free all it's memory, close all it's file descriptors, but keep
> the state around as a zombie for tracking IO completions or whatever.
Yeah. As said, practically there isn't much difference, so I assume noone
bothered, And as Marcus said, the kernel should not block for very long
times anyway ...
> >> On the other hand, whatever is doing the IO can tell how long something
> >You mean the application ? No. I don't think it has an idea.
> No, the application doesn't do IO. The application requests IO.
O.K. - simple misunderstanding here.
> I recall -- they made no distinction between a STREAMS source
> in the kernel or in user space.
*grin* sounds very much like EvStack :-).
> Yea, well, they all write to Microsoft's standard programming model ...
Which is ? Display welcome screen, wait for 25 keypresses or mouseclicks,
then crash ? :-)
CU, ANdy
--
= Andreas Beck | Email : <[EMAIL PROTECTED]> =
