Soren S. Jorvang writes:
 > I don't know of anything similar done for NetBSD, but given
 > the many things that are outside your control, polling the
 > ctime of the incoming directory seems like a comparatively
 > small evil to me.

I'm certain I'll end up doing some sort of userland thing like a
semaphore rather than polling.  Polling is a smallish evil, but CPU
cycles *really* matter in this case.  Also, using the "p" work around
here tends to get you boo'ed.  Of course, I end up beating every cycle
I can out of the machine, then someone else puts in something like
this:
system ("mv file file.old &"); 
sleep (10); 
which more than makes up for the few cycles I saved.  It always works
out that way... :-(

 > Out of interest, what sort of products are you planning
 > to use NetBSD in?

If I work on this product, probably prototyping only.  The previous
software ran on Solaris.  They'll probably stick with Solaris unless
there is a compelling reason not to.  Money not being an object, I
doubt NetBSD will make the final cut, but since I'm more familiar with
NetBSD than Solaris, if I do work on the project I would probably
start with NetBSD then port to Solaris.

-- 
Eric McWhorter
Xerox Special Information Systems
emcwhorter@xsis.xerox.com