Re: Re: Adler-32 Checksum

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Friday 01 July 2005 16:32, James Carlson wrote:
> Thomas Dreibholz writes:
> > On Thursday 30 June 2005 17:06, Peter Lei wrote:
> > > I'm not sure what your concern is here.  The Internet checksum
> > > is just the ones's complement of (one's complement sum + carry).
> > > That's two additional steps over a simple XOR sum.
> >
> > XOR is computationally cheaper than addition. It works bit-wise and does
> > not need a carry.
>
> Was that meant as a serious comment?  I don't think that sort of
> consideration applies to _any_ system built in the past 40 years or
> so.

XOR is cheaper, although not significantly on today's computers.

> So if the probability of detecting handlespace inconsistencies 
> would be the same for the Internet checksum and XOR, XOR would be the
> preferable solution. 

But unfortunately, the inconsistency detection probability seems to be lower
for XOR. Example:
Pool "TEST", PE #1000
Pool "TEST", PE #1001
Pool "TEST", PE #1002
Pool "TEST", PE #1003
Since the amount of PEs in the pool is even, the PHs' XOR chechsum would
cancel out (x XOR x XOR x XOR x == 0); therefore the checksum would
effectively be computed over the PE IDs only.

Using the Internet checksum, the PHs - regardless of even or odd number of
PEs in a pool - would always contribute to the checksum. Therefore, the
Internet checksum is the preferred solution.


Best regards
- -- 
=======================================================================
 Dipl.-Inform. Thomas Dreibholz

 University of Essen,                            Room ES210
 Inst. for Experimental Mathematics              Ellernstraße 29
 Computer Networking Technology Group            D-45326 Essen/Germany
- -----------------------------------------------------------------------
 E-Mail:     [email protected]
 Homepage:   http://www.exp-math.uni-essen.de/~dreibh
=======================================================================
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFCyPRv32BbsHYPLWURAnfdAJ4xAF/mcRmFZ4h1HEPdRANKuWkJmQCePrdg
1GxKWH8tQgEQDdYoypNhTFU=
=UgQJ
-----END PGP SIGNATURE-----