Re: Wired article on Jahn

[Stephen A. Lawrence]

Jed Rothwell wrote:

Grimer wrote:

I have to respectfully disagree with that statement. Suppose you observe
some scientific phenomena which only occurs once and you are the only 
observer.


If it only occurs once then it isn't scientific -- yet. You have to 
reproduce it. If you cannot reproduce it, then eventually you must 
conclude that you did not see it.

Nonsense.  If you cannot reproduce it you must conclude that it is very 
hard to reproduce.  There's no need to conclude that you did not see it 
at all (unless, of course, you weren't quite sure you saw it to start with).

Speaking as a programmer, we encounter "Can Not Reproduce" situations 
all the time.  In fact, any reasonably complex program is almost certain 
to have bugs that occur so rarely that they will be nearly impossible to 
reproduce just because the frequency of occurrence is so low, so even if 
you see it once, you'll probably never see it again.

Speaking as a hardware designer, modern machines have numerous circuits 
that work "most of the time" embedded in them; if they're properly 
designed, "most of the time" means so much of the time that nobody will 
ever see them fail.  A classic example is edge resynchronization when a 
signal crosses from one clock domain to another.  If you try to latch a 
level from a signal that came from another clock domain, sometimes your 
flip flop will fail to settle down in the allotted time, and your 
machine will fail.  The probability of this happening is calculable and 
depends on the properties of the signal and the flipflop, and one 
normally tries to make sure it won't happen more often than, say, 1 time 
in 1 trillion years of operation.   So, suppose some user actually sees 
a crash caused by such a synchronization failure -- we can be very, very 
sure that it's an event that will never happen again, for all our usual 
definitions of "never".  But that doesn't mean it didn't happen _once_!

Computers are complex, but so is nature, and so are human brains.   As I 
just said, I know for a fact that "CNR" events can happen in computers.  
Why would you feel they can't happen in any other arena?  If I observe 
something I'm sure is a "CNR" event in nature, I may conclude that I 
should tell nobody about it because I won't be believed, but it is not 
rational to say that I must eventually conclude that I did not see it!