Non broadcasting rwhod

[PREVELAKIS Vassilis]

In article <3674 at phri.UUCP> roy at phri.UUCP writes:
>(*) Just why was I hacking on rwhod you ask?  As documented by Sun, rwho is a
>real performance pig. [...]

You came 3rd :-)  About three years ago, I changed rwhod so that all the useless
broadcasting could be eliminated.  When I came here (to Geneva) I found out
that another guy here had also performed (more or less) the same mods on
rwhod.

The way I did it, was as follows:
	If rwhod realizes thats its running on a client machine it simply
	sends its load info to the machine designated as server.

	The rwhod running on the server simply collects the load info
	from the clients and updates files in the /usr/spool/rwho
	directory (it also saves there its own load info).

	The client machines NFS mount the server's /usr/spool/rwho
	directory so that programs like ruptime work without mods.

	How does rhod know who is the server?  The easy way is to
	supply it as an argument (e.g. /usr/etc/rwhod <servername>).
	(or for that matter have /usr/etc/rwhod linked to
	/usr/spool/servername and have rwhod check its own name)
	Anyway, this gives one version of the rwhod program without
	causing too much hassle in case you want to designate
	a new rwhod server.

Probs:
	I also got bitten by the special version of sendto.  For some
	(unknown) reason I assumed that the debugging version printed
	the diagnostics but SENT THE MESSAGE AS WELL.  Needless to say
	that I was very annoyed when I realized that I was looking
	for bugs that weren't there.

	The other problem occurs when the clocks of the various machines
	are out of sync.  In that case the server may get times in the
	future and assume that the machines are down.  Someone told me
	that this is no longer a problem, but I am still skeptical.

BTW some guy was asking about broadcasts over the Ethernet.  I think that
the rwhod code is good example.	 I also think that the rwhod code is in the
stuff that can be ftp'd from UCB free of charge.
	
/vp