CD Ripping Uncertainty Again, Update

An update on the issue of the reliability of cdparanoia rips (on Mac OS 
X), posted 2006/03/23.

Since then, I ripped around 50 CDs, in various conditions, some new, 
some used pretty roughly, using cdparanoia on Mac OS X v10.3.9 with a 
Plextor Premium drive. Each time I ripped every track twice and 
compared MD5 checksums. (I automated slightly the process using 
AppleScript.)

In most cases, the MD5 checksums matched, and the rips were good (I 
tested some with EAC on Wintel machine). In a few instances, the MD5 
sums didn't match. I tested those tracks with EAC on a Wintel with a 
Teac DE224 drive. A few EAC was able to extract, but with extensive 
error correction; for others, EAC reported read error/sync error. All 
of them were on old, scratched library CDs.

There was one exception, a new CD from EMI with the Berliner 
Philharmoniker under Sir Simon Rattle performing Debussy's La Mer and 
other pieces.

<http://tinyurl.com/nqeel>

On this CD, EAC failed on track 1 as soon as it started; cdparanoia 
ripped it without reporting problems, but the MD5 sums didn't match (I 
did about 10 rips; the checksums matched for the other tracks). I 
compared the WAVEs with EAC, and there were on average 5 difference 
between the rips (two rips differed by a single sample). However, I 
wasn't able to hear any difference, and, as far as my ears could tell, 
all rips were good.

There wasn't a single instance when the checksums matched, but the rip 
was audibly bad.

My conclusion is that, at least with my setup, ripping twice and 
comparing checksums is a fairly reliable method of ensuring that 
cdparanoia rips are good.

thanx