"Dr. Imre Bartfai" wrote:

> only as an idea:
> I am also interested in fingerprint identifying. If I'd do it, I'd store only
> the e. g. 32-bit CRC of the fingerprint to be identified. It hashes pretty goo
d
> (2^32 is about 4*10^9).

Auch! No it doesn't.
Try fingerprinting with CRCs and you'll find a heck of a lot of unknown people
with valid access ;-)
I wrote the following text about six years ago....
[btw, the use of the word 'fingerprint' in that text refers to uniquely
identifing a large set of data by a smaller 'digital fingerprint' - but of cours
e
the same principle applies if you have a human-fingerprint reader. It's a set of
data.....]
Regards, Arjen.

  WHEN NOT TO USE CRC

  Lately increasingly weird applications of CRCs have popped up, written by
  people who don't understand CRCs and their limitations.
  The CRC calculated from a block of data is NOT unique; it is highly probable
  that a different block of data gives the same CRC value.
  Many people respond with "yes, but do look so unique".
  They may well look unique to us, but if you need a unique fingerprint to
  compare two blocks of data, CRC is NOT the right method.
  CRCs can identify certain kinds of bit-errors in a limited length data block,
  not distinguish two or more completely different sets of data!

  Mis-application of CRCs in this way has caused heaps of trouble in for
  instance FidoNet... One program used CRCs to distinguish newsgroup names,
  the result was a load of crosslinks. CRCs were also used for detecting
  duplicate messages; the result was that good different messages were branded
  as duplicates and therefore deleted.
  All this was NOT a result of bad implementation, they were just matters
  where an algorithm was used for something it wasn't suitable for.
  Please don't make the same mistake, the users deserve better.....

  For fingerprint purposes I strongly advise using the Message Digest algorithm
  (MD4 and now MD5) by Donald Rivest. Specs and sources are freely available.
  If you can't find them, contact me.
  If you have a FidoNet mailer, you can file-request MD.ARC from my system,
  that file contains the original specs for both MD4 and MD5, portable C/C++
  sources, and validation/speed-check programs.
  BUT WHATEVER YOU DO, DON'T MIS-USE CRCS!!!