Unfortunately, I didn't know much about cd ripping, and after ripping my cd collection using the basic ripper, I started noticing some pops, which led me to reading about secure ripping and the difficulties thereof on your forums and elsewhere.
My first suggestion would be to not make the basic ripper the default so noobs like me don't shoot themselves in the foot
On to the interesting part, semi-cross posted from my blog:
From what I understand, secure rippers such as cdparanoia and the comparison ripper basically read each sector multiple times until it gets the same data from a particular sector n times. This does a decent job of getting rid of non-deterministic read errors. Unfortunately, cd's can be damaged in such a way that they consistently return the same incorrect data, so it's not foolproof. And of course if a sector is horribly damaged you may not be able to read anything at all.
That led me to wonder if there would be value in a collaborative ripping database. If, after ripping a cd, people submitted hashes for each track, and/or each sector, we could compare the hashes across multiple discs, and have a much higher confidence of whether the correct data was read. This would at least let you know whether you successfully ripped a sector/track/disc, but if the disc is damaged such that you can never read the correct data, you're still screwed.
Even better would be if the database could *correct* bad sectors. A naive solution would be to store the actual audio data, and provide requested sectors to people. Of course, this would be heavy weight, highly abused, and slapped with a cease-and-desist before you could say fair-use.
One way that might work is for people to generate and submit error-correcting codes per sector (e.g., reed-solomon). These codes could be used to correct a sector with a relatively small number of bit errors, but wouldn't give much useful information to someone who didn't have the original (maybe slightly damaged) sector.
Any thoughts on this? My usual expectation when I think of something like this is that either it's been done, or there's a good reason why it hasn't been done. If nothing else, it would require a server with pretty decent storage and bandwidth.
EDIT: btw, sorry if this is off-topic. Feel free to move it to the 'general discussion' forum if it belongs there. I posted here because, while I forgot to explicitly ask, part of what I want to know is if such a service existed, would you be interested in integrating it into max?
