Compression on CDs

[Loye Young]

On Thursday, October 25, 2007 5:46:01 pm Soren Hansen wrote:
> I'm quite sure this is not true. The burning process and the error
> correction mechanisms employed on compact disks is complete agnostic
> towards the nature of the data stored on the medium. From anything below
> the application level, the data on the CD is just ones and zeroes (and
> the occasional out-of-band data such as track boundaries) whose
> likelyhood of read or write error is the same regardless of the entropy
> of the data stream it's a part of.

Soren,

What you say is true as far as it goes, but the issue isn't whether there are 
more read/write errors, but instead whether the read/write errors that do 
occur actually make a difference.

Think of it like meteorites hitting the earth. Gene Shoemaker of the US 
Geological Society estimated that an impact event with a force of the 
Hiroshima bomb hits the Earth approximately once a year. 
http://www.vectorsite.net/taimpact.html. But fortunately, most of the world 
is covered by oceans and wilderness lands, so the vast majority of all 
meteorites that fall from the sky fall harmlessly into the sea or on places 
nobody cares about (like Arkansas, USA, where the residents call them UFOs 
and blame the French, the Russians, or the Pope). See 
http://en.wikipedia.org/wiki/Impact_event#Modern_impact_events. So, apart 
from the occasional sci-fi thriller, meteorites don't cause much damage. If, 
however, the whole world were covered with land that were as densely 
populated as Tokyo or New York City, meteorites would be of greater concern. 

Similarly, if data on a disk is not compressed, large areas on the medium are 
essentially empty, so read or write errors, while they may occur, don't often 
affect data that's important. Compression, on the other hand, more 
efficiently uses that space, so more of the surface area actually has data. 
Thus, the same number of read/write errors are relatively more likely to 
affect important data. The higher the compression, the greater the density of 
important data and the greater likelihood that read/write errors will cause 
failure. 

Apart from theoretical analysis, I observe (as least anecdotally) that disks 
with compressed data have higher failure rates than with uncompressed data. I 
don't think I'm alone in my observations, but I admit that I haven't 
conducted rigorous scientifically tests, either. 

Happy Trails, my friend.

Loye Young