Re: [MiNT] Hash filename for index

["Guido Flohr" <gufl0000@stud.uni-sb.de>]

On Fri, Jun 23, 2000 at 09:17:06AM +0200, Julian Reschke wrote:
> It certainly is *much* slower than a plain 32 Bit CRC (POSIX cksum).
> 
> BTW: when I wrote the code that is in TOSFS, I checked ir against all my
> partitions and found no duplicates.

> [ and Thomas suggested MD5/SHA-1. ]

I think we should really test how these three algorithms behave on the
expected input, i. e. relatively short strings with a very restricted code
set (mainly lowercase US-ASCII plus dot, slash/backslash).  If CRC32
doesn't produce too many conflicts it is of course preferable because the
hash algorithm doesn't have to be cryptographically secure.

I am convinced that CRC32 will really do but I will test in on an average
file system, too.  My linux box has a lot more files than my TT (and
enough RAM to keep the counter tables).

Ciao

Guido
-- 
http://www.stud.uni-saarland.de/
Send your spam to president@whitehouse.gov and your replies to
mailto:guido at freemint dot de