[gclist] Re: gclist-digest V2 #76

Giuliano Carlini GCARLINI@us.oracle.com
10 Dec 97 19:39:08 -0800

Content-Type:text/plain; charset="us-ascii"

RE: I would be very careful about analyses like this.

Whole heartedly yes!

I meant only to explain that while Mike's comment wasn't
literally true, it was a close approximation. I relied
on peoples good sense to realize that approximations are
just that.

By no means should you do anything where relying on the
statistical approximation of precision would cause an error,
for example a copying collector. After all, 1 in a million
multiplied by a million users a day means one crash a day.
That would suck if the app crashing would kill someone.



Date: 10 Dec 97 17:15:39
From:John R Levine <johnl@iecc.com>
To:Giuliano Carlini <GCARLINI@us.oracle.com>
Subject:Re: [gclist] Re: gclist-digest V2 #76
Precedence: bulk
MIME-Version: 1.0
Content-Type:text/plain; charset="US-ASCII"

> There is a very small chance that a random integer will
> correspond to a block. Consider a 2^N heap all of whose
> data are uniformly distributed integers. The chance of any
> 1 integer corresponding to some block is 2^(N-64). ...

I would be very careful about analyses like this.

I have this really great 64 bit processor chip, with a minor bug in the
"compare" instruction so it will never say that its two operands are equal. 
But that doesn't matter, since the chances of two uniformly distributed 64
bit numbers being equal is 2**-64 -- if a program running on the chip
executed one comparison every 100 nanoseconds, statistically it would take
over 29,000 years* before a pair of operands were equal. 

John Levine, johnl@iecc.com, Primary Perpetrator of "The Internet for
Information Superhighwayman wanna-be, http://iecc.com/johnl, Sewer
Finger for PGP key, f'print = 3A 5B D0 3F D9 A0 6A A4  2D AC 1E 9E A6 36 A3

* - actual number