If anyone has any thing to add to this, please let me know.


o	capture first n seconds of sucessful connects. tickle connection
	with bytes from user specified file. display with reg stats.

o	when we hit 111 slurp port mapper descriptions. map to description
	&& hit lists. hack up some portable non-blocking rpc routines, or
        at least hand craft the required sequence.

o	make the port list more oo. have a func and exec field. pass critical
	data into env, sock to stdin/out. fork off seconder func/exec. take
        stderr and add to port desc. need to limit forks to reasonable 
        levels. don't loose sight of speed/paralell objectives.

o	hack up a portable simple threads package, so we can get out of being
	a parallel finite-state machine. Can we do this effectively by using
        setjmp to pass control back to the threader at non-critical moments?

o	add checking for ICMP_ECHO's via SOCK_RAW. merge into an exclusive
        pre-scan. this really should *fly*, should be able to do about 2k
        hosts a second on a T1, under optimal conditions ping every host in
        the world in < a hour. consider record route, given hop limitation?

o	write output parser to beautify/segment output according to 
        machine/domain/topology.

o	investigate hand crafting (via /dev/nit etc) the initial TCP con
        req packet, and pulling the result out directly. we avoid kernel
        socket limitations and sending first ACK. should be able to
        avoid all tcp wrapper (etc) logs this way, as well as save on
        bandwidth. Is this worth the portability/root perms issues?

o	include simple udp checks for well-known services. maintain external
        db of snached udp packets, w/pointer -> reply/send addr offsets. user
        configurable retries. investigate adding target machines to a multi
        cast group.