>types of numbers you're working with, and you're writing to CD-R or DVD-R
>media, and you're replacing/refreshing your data once every 5 years,
>you're going to go through 15 cycles.  And the process is NOT
>"lossless".  No one has come up with exact figures as to how much loss is
>acceptable, but it's not uncommon to experience 1-2% loss per migration...


Larry, I'll put together the economic analysis for you in a week or two,
but in the meantime, I couldn't let this go unchallenged (despite your
other good points.)  It is inaccurate to state that digital conversion is
not lossless and grossly inaccurate to say that there is a "1-2% loss per
migration".

If you use CD's  - which are less reliable than a standard magnetic disk
but more reliable than non-DAT magnetic tape over short terms - there are
two types of redundancy built into the CD.  I won't bore you with the math
and details of how it works, but the *uncorrected* bit error rate is about
1 in 10^(-12) or one per trillion bits.  An 8.5x11" 200 dpi image should
have 3.74M pixels (raw).  Some of our older research suggests than only
about 29% of the image contains "significant" information (i.e.,
information used for actual reader interpretation).  The numbers may have
changed a bit, but it's probably still in the ballpark.  Let's be generous
and say that 0.1% of the pixels could change an interpretation.  The math
works out pretty even so that the changes of a single meaningful pixel
change are just under 1 in 10 trillion.  (Want some perspective?  The
estimate for number of stars in the universe is 10^(11) or ten times less
than the error rate.)  Hardly a 1-2% loss.

List archives at http://lists.ufl.edu/archives/recmgmt-l.html
Contact [log in to unmask] for assistance