Infant mortality and longevity of DVD media?

Jerome H. Fine jhfinedp3k at
Wed Oct 10 19:22:17 CDT 2007

 >Jim Leonard wrote:

> >Jerome H. Fine wrote:
>> But I have found
>> that with a 1 GB file, it takes less overall time to copy the file to 
>> a hard drive
>> first.
> This is a buffering issue with your md5 binary, not DVD_R technology 
> itself.

I agree.  On the other hand, it is painful to wait many minutes for each 
1 GB file
to be checked (produce the MD5 value) while still on the DVD and then 
check the next 1 GB file.  So even if the time was the same to copy or 
check the
file, it is far more convenient to just copy all the files on the DVD to 
the hard drive
at which point I have also verified I can read the complete file.  If I 
then check the
file copied from the DVD media (reproduce the MD5 value), I am confident 
the file on the DVD is completely valid.  Interestingly, reproducing the 
MD5 value
from the file on the hard drive takes just under a minute (just about 
the read speed
of the file) and it takes me about the same time to manually scan the 
original saved
MD5 value and the new value and compare them.  And I just discovered 
that the
length of the MD5 value is 128 bits (32 hex digits - I had though it was 
that long
but somehow the 32 hex digits always became 32 bits when I though about 
the size)
which should be more than sufficient to confirm that the file I have 
read with the same
size as the original is IDENTICAL to the original.

I remember that Tim Shoppa used MD5 values for BZ2 files that were 
Since it is rare for downloads to be corrupted, MD5 is probably the most 
efficient way
to verify that the downloaded file is identical.

I agree that it would be better to use some methods to restore the 
archived file if
corruption takes place, but yearly testing should minimize that 
problem.  Plus with
a monthly backup image, the file I want to recover is most often on more 
than a
single backup image.

Sincerely yours,

Jerome Fine

More information about the cctalk mailing list