🚀 go-pugleaf

In article <4f6bag$ckp@news.mpd.tandem.com>, Joseph Crowe <jcrowe> wrote:

>    Here you hit on the one essential issue that probably causes all of the
> conflict in the PMT vs. CMT debate.  Essentially, with PMT, the OS controls
> what the app can do, such that the writer of the app should not have the
> capability of "taking over the machine". 

But we give that right up all the time.  Heck, it is trivial to do a
denial of service attack to a host on a TCP/IP network.  Again, in
practice, it doesn't come up very often.  Since it is indeed possible,
should we give up our networks?

> >That is because one of the rules of programming a CMT system is that you
> >have to give up time for other processes to run.  Duh.  Are you really,
> >truly, honestly so dumb that you don't understand this?  If you really
> >don't, then you are too stupid to be programming a Mac or Windows or DOS
> >machine.

>   You're reaching...the telltale sign is that you feel it necessary to insult
> the other individual in the debate rather than apply reason and logic.

No.  There are some people you can't convince, no matter how much reason
and logic you use.  Example:  the Monty Hall problem on rec.puzzles (no,
don't discuss it here and for their sake, don't dicuss it there).  Some
people believe that the answer is 50%, not 66 2/3%, and that debate has
been going on longer than this one!  It doesn't matter how well you
present the correct solution to them, they are just too stubborn or too
dumb to get it.

The debate has gone something like this:

Don:  In a CMT environment, a non-cooperating process can starve other
processes.
Me:  You bet it can.  In practice, almost none of the applications exhibit
this behavior.
Dan:  But in a CMT environment, a non-cooperating process can starve other
processes.
Me:  Again, you haven't told us why you think this is a big deal.  Is
there some evidence that this really does occur in practice?
Kevin:  Except in a CMT environment, where a non-cooperating process can
starve other processes.  Here is another contrived example of what can
possibly do this.
Me:  Trivial to fix.  Who cares?  It still doesn't happen in the stuff we run.
Don:  Except in a CMT environment, where a non-cooperating process can
starve other processes.
Me:  Yup.  Then again, there are better selling points for PMT than the
ability to run crappy software.
Dan:  However, in a CMT environment, it is possible for a non-cooperating
process to starve other processes.
Me:  Well, Mac programmers must be an incredibly smart bunch, since for
some magical reason, this particular problem just doesn't seem to happen
on Macs using real software that people buy and sell.
Don:  You do know that in a CMT environment, it is possible for a
non-cooperating process to starve other processes.

At this point, I have to question their ability to reason.  I guess they
think the issue will magically become important if they keep repeating it
over and over and over and over and over and over and over.  Either they
just like to waste bandwidth (beer commercials made fun of these people in
the "Tastes Great"/"Less Filling" ads), they don't have the intelligence
to figure out that this is an incredibly minor and trivial point, or I
don't have the intelligence to realize why this is so important that the
100 million of us that don't have it really can't live without it, even
though we are living without it.  It seems so obvious to me that processes
that don't cooperate in a cooperative environment can do bad things, and
applications that do bad things tend not to get used in favor of those
that don't do bad things.

Now, a stronger person than myself would bow out of these debates at this
time.  I not strong enough to do so, but I do take comfort in the fact
that I am no weaker than Don or Dan or Kevin.

>    Yeah, but not all.  Think about this....you know all of those annoying
> alerts that essentially allow the user no alternative but to respond before
> doing anything else on the Mac?

That is a human interface problem, not a multitasking problem.  The other
apps still get time when a modal dialog is up; you just can't respond with
human actions (clicking the mouse, etc.) for those other processes.  It
isn't good to do this, but it is a vastly different issue than CMT vs PMT.
-- 
        Nevin ":-)" Liber       nevin@CS.Arizona.EDU    (520) 293-2799