md5sum gives odd results. . .

I was downloading Knoppix, and checked it against the md5sum, and it didn't 
check.

So I downloaded it again, from a different mirror.

Then I realized that md5sum was giving me different results in different runs.

It didn't matter whether I was using the -t flag or the -b flag, it would 
alternate between one result and the other, for a given file, which is really 
weird.

Since /neither/ of the checksums it was producing matched the Knoppix ( 
LiveCD Linux Distro ), iso, I deleted that and am downloading it again ( the 
two I'd downloaded /were/ different, and with wget they shoudn't be, and 
neither checked out ).

In trying this on the "Beehive" linux iso, I find that md5sum gives exactly 
predictable results, 
4 times with -t, 4 times with -b, same checksum?
f3639773ddc448f9b39776f7d0fc1ff6  beehive_linux-0.5.0.iso
f3639773ddc448f9b39776f7d0fc1ff6  beehive_linux-0.5.0.iso
f3639773ddc448f9b39776f7d0fc1ff6  beehive_linux-0.5.0.iso
f3639773ddc448f9b39776f7d0fc1ff6  beehive_linux-0.5.0.iso
f3639773ddc448f9b39776f7d0fc1ff6 *beehive_linux-0.5.0.iso
f3639773ddc448f9b39776f7d0fc1ff6 *beehive_linux-0.5.0.iso
f3639773ddc448f9b39776f7d0fc1ff6 *beehive_linux-0.5.0.iso
f3639773ddc448f9b39776f7d0fc1ff6 *beehive_linux-0.5.0.iso

yet, the same program ( the latest, compiled from source, against a 2.4.19 
kernel ) gives, on the Knoppix iso, this ( cat knoppix.md5 first, then 
"md5sum -b" 4 times, then "md5sum -t" 4 times )

9222fe7f69bb1b5f3ab39758e9b471cd  KNOPPIX_V3.1-04-08-2002-EN.iso

9222fe7f69bb1b5f3ab39758e9b471cd *KNOPPIX_V3.1-04-08-2002-EN.iso
a951659a28666e00fa6b2b9ed0c3b16d *KNOPPIX_V3.1-04-08-2002-EN.iso
f4257710cc0c93044c0562f16c766fa3 *KNOPPIX_V3.1-04-08-2002-EN.iso
a951659a28666e00fa6b2b9ed0c3b16d *KNOPPIX_V3.1-04-08-2002-EN.iso
0e4d8925327bf4bee30efc6c4ea05e7b  KNOPPIX_V3.1-04-08-2002-EN.iso
f42cf3be02e0b329aee97dcfbc9ab84e  KNOPPIX_V3.1-04-08-2002-EN.iso
a58d70e51d9f9e90c372bc574b04e5f1  KNOPPIX_V3.1-04-08-2002-EN.iso
0e4d8925327bf4bee30efc6c4ea05e7b  KNOPPIX_V3.1-04-08-2002-EN.iso

Dig That.
1 match, 2 a95's 2 f42's 2 0e's and an a58, on a stacked command that did the 
same md5sum . . . 4 times in a row each way ( no interaction while this was 
going on )

Wait, here's something weird, when I burn the ISO, then "diff -r" it, I get
Binary files   \
/mnt/loopdisk/KNOPPIX/KNOPPIX and   \
/mnt/cdrom/KNOPPIX/KNOPPIX differ

so I glance at 'em ( in "less" ), and they're links to the entire image?
a "file" on the thing that loops to the entire image, eh?

MAYBE that's what's weird about this file that I can't md5sum it correctly.
I /think/ it is a loop to the entire image ( more or less ) simply because 
the singel "file" is over 700MB large, and that's almost the complete size of 
the iso.

Right, so now I don't know if there's a bug in md5sum ( that it /should/ deal 
with strange cases like this in some informative way, at least stating that 
this is going on. . . ), or if this is a case that /cannot/ be dealt with in 
any way that would inform the users of md5sum what was going on or why.

<sigh>  what a waste of bandwidth, me downloading the image 3 times to get 
this far. . .

Right,  Here's the info I scrounged, here's the wget.list file I used:

ftp://ftp.tu-chemnitz.de/pub/linux/knoppix/KNOPPIX_V3.1-04-08-2002-EN.iso.md5
ftp://ftp.tu-chemnitz.de/pub/linux/knoppix/KNOPPIX_V3.1-04-08-2002-EN.iso
ftp://ftp.tu-chemnitz.de/pub/linux/knoppix/README
ftp://ftp.tu-chemnitz.de/pub/linux/knoppix/packages.txt
ftp://ftp.uni-kl.de/pub/linux/knoppix/KNOPPIX_V3.1-04-08-2002-EN.iso.md5
ftp://ftp.uni-kl.de/pub/linux/knoppix/KNOPPIX_V3.1-04-08-2002-EN.iso
ftp://ftp.uni-kl.de/pub/linux/knoppix/packages.txt
ftp://ftp.uni-kl.de/pub/linux/knoppix/README

/you/ try figuring it out.