[net.general] How many readers ARE there???...

ignatz (11/30/82)

	You know, a while back there was blathering about just how many
computers are on USENET; and comments like, "...remember that what you say
is read by thousands of people..."
	Really?  I've never noticed thousands of signatures; not even
hundreds.  Has anyone ever been able, either as a formal study or an informal
survey, to truly determine the size of the USENET community?  It would
have to include all passive readers, but I think it could safely be limited
to, say, only readers who have the "direct capability" of participating in
the conversation.  Thus, the fact that by fanning out of one gateway, an
article picks up N-thousand ARPANET readers would be ignored in the actual
count.
	One way that came to mind immediately was to just say, "Everyone
who gets net.general, just pop a message in the mail back to me.  I'll count
and report...".  Except that, if there ARE thousands of readers capable of
replying, I'd be performing another study, concerning overload of a single
Vax Unix system.  Please note that, unless a local SA sez to try it, this
is NOT being advocated at this time.
	Ideas? Questions?

		A Curious one of {thousands|hundreds|dozens|a few}
		(Select the appropriate range, if you know),

		Dave Ihnat
		ihuxx!ignatz

Onyx:harymudd (12/02/82)

	The method you mentioned of overloading a VAX UNIX doesn't work very
well.  The best way is a process that forks itself (somebody during October
did that to Vax Populi thru unauthorized use of my account).  The second best
way, although I haven't tried it on UNIX yet since it doesn't have the
software I used, is to have a process with loads of I/O.  Using last year's
version of Datatrieve (Certain people at test sites who are reading this
should know about this), I was able to bring a VAX 11/780 using a five
factor cross command.  That was fun...
	One of the more popular ways to slow down a VAX UNIX in Berkeley is
to load many students on it.  Fourty people running the pascal compilier at
once tends to do it.
	One note to your posting is that in Berkeley on the major class
machines, readnews is suppose to be found, not known about.
	Live Long and Prosper, and May the Force Be With You...

trt (12/03/82)

My favorite way to crunch UNIX is with a cpu-bound non-blocking system call.
E.g. UNIX-V7 has a slow mem(IV) driver, so "dd if=/dev/kmem bs=50b" is SLOW.
Such things cannot be preempted, so the system looks quite dead.
Opening the file "/etc/../etc/../etc/...on and on" is just as bad.
The following will make a VAX 780 autistic for about a minute:
	char	huge[250000];
	main()
	{
		register char *p = huge;
		register int n = sizeof(huge)/2 - 10;

		do {
			*p++ = '.'; *p++ = '/';
		} while (--n);
		*p = '\0';
		open(huge, 0);	/* Be patient! */
	}
Do not run the above program!! You might lose your login privilege.
Perhaps UNIX should defeat such calls by calling swtch() occasionally.
	Tom Truscott