kick ass idea to make you super famous

[groovyd]

one kick ass idea that would make your ripper super bad ass is to iterate all combinations on bad frames until you find the one that gives the 'accurate' CRC. Granted this could take some time but it would be the ultimate in error correction.

ofcourse it is only practical for the case where you have a single bad frame or two, and you know which frame is bad.

:teufel8:

[groovyd]

Re: kick ass idea to make you super famous

If you also kept a straight running checksum of each frame then you could assume the difference between the accurate checksum and the faulty one is the error in one sample and then apply that to each sample checking then if the CRC matches. If one sample doesn't do it then you divide the error across 2 samples and try again, then 3... the combinations grow exponentially with the number of samples in error, but it would be interesting to see if it catches anything for low numbers of samples...

I guess the question is normally how many samples in a bad frame are bad and can this even be determined during the rip process? If the answer is under a dozen this would work well.

[MedO]

Re: kick ass idea to make you super famous

There are 6.3*10^57 possible values for a frame. It is impossible to iterate them all in the lifetime of the universe, except if you have a quantum computer or something. If you mean that it should try all combinations of actually ripped data, that *might* be feasible, but I can still see several problems with that, not the least of which is that AccurateRip-Checksums might be too short to still guarantee then that the data will actually be accurate if used like that.

I'd propose a different way, but that would blow up the AR database quite a bit, and would need more calculation time for AR submissions, at least for CDs not yet in the database (not to mention possible legal issues, but that might not be a problem): Create Reed-Solomon ECC blocks for the disks and store them in the AccurateRip DB. Let's say you want to enable people to be able to still rip a CD accurately if they have up to ~400 bad frames, you would need 10kb extra data in the DB for every disk. If the ripper runs into less than 400 bad frames, it will mark them as erasures for the ECC, apply the correction and check with the AR sums already in the DB.

I doubt anything like that will be implemented though.

[EliC]

Re: kick ass idea to make you super famous

MedO, I have actually already proposed exactly this idea. I think it would be fantastic!!! I don't know the specifics of how something like this would be setup but Yahoo hosing now provides for unlimited space and bandwidth for only $12/m.

[Spoon]

Re: kick ass idea to make you super famous

There is no such thing as unlimited bandwidth, once you push over 1000 Gig a month you would get the boot.

[EliC]

Re: kick ass idea to make you super famous

That may be, but I would be VERY willing to pay a premium, even a per-lookup fee for this service.

[bhoar]

Re: kick ass idea to make you super famous

That may be, but I would be VERY willing to pay a premium, even a per-lookup fee for this service.

There's also a legal liability issue that spoon brought up the last time I talked to him about making a parity-type 'correction' service available. Single checksums are one thing, but parity data potentially gets him into the realm of "distributing content"...at least when dealing with existence of hyper-litigious copyright control organizations...even if I disagree with that assessment, that's how they'll see it.

-brendan

[EliC]

Re: kick ass idea to make you super famous

I can understand why spoon would not want to deal with it, since they have bottomless pockets, but CLEARLY there is no content. Show me a player that would be able to play back even a single note with the "content"

[LtData]

Re: kick ass idea to make you super famous

You're attempting to use logic in an area where the RIAA is, you should know better than that, EliC :P

[EliC]

Re: kick ass idea to make you super famous

here is an earlier thread:

http://forum.dbpoweramp.com/showthread.php?t=10945

[Porcus]

Re: kick ass idea to make you super famous

Wikipedia soundclips are constrained to 10 % of the length -- and 64 kbps, which at worst would be ~ 1/20 of the total content. Total: Half a percent.

So if the error codes are < .5% of the track (check, 10k < track data / 200)
*and*
at "sound quality which is lesser than 64 kbps" (check, it would be useless to play) --
OK, throw in:
*and*
the software does not retrieve error codes for more than 10 % of the length

-- then we would be well off, except that the "10 %" can vary?

You're attempting to use logic in an area where the RIAA is, you should know better than that, EliC :P

Ouch! Relocate outside the US then ;-) Ask DVD-Jon to where.
(Hm: Has CDDB or freedb had any issues? They are distributing meta-content too.)

There is no such thing as unlimited bandwidth, once you push over 1000 Gig a month you would get the boot.

Use BitTorrent. Safest way to get your ass sued :D

[bhoar]

Re: kick ass idea to make you super famous

...the software does not retrieve error codes for more than 10 % of the length...

Ouch! Relocate outside the US then ;-) Ask DVD-Jon to where.
(Hm: Has CDDB or freedb had any issues? They are distributing meta-content too.)

Spoon runs a business. DVD-Jon does not. Different liability model.

Let me put it this way: is it worth if spoon implemented this type of feature and then an injunction was put in place that essentially destroyed his business and he had to stop working on dbpa entirely?

The risk/reward number is way too high.

Meta-data in general is presumably covered under fair-use (it's a list of content, not the content itself), which is why this hasn't been a problem (e.g. why libaries do not get in trouble for card catalogs and why Amazon.com can list all the chapters in a book, etc.).

-brendan

[Porcus]

Re: kick ass idea to make you super famous

Let me put it this way: is it worth if spoon implemented this type of feature and then an injunction was put in place that essentially destroyed his business and he had to stop working on dbpa entirely?

Nope. So let someone else in a less hysterical jurisdiction do the dirty work of hosting the server non-profit. And leave open for a "test and check" module which defaults to AccurateRip lookup, but customizeable by the sourceforgers.

Meta-data in general is presumably covered under fair-use (it's a list of content, not the content itself)

But are error correction codes metadata? Parity protection bits might not be -- if you put up a RAID array of 3 HDs, then they are equivalent and each are as much data as another. But a checksum? A checksum for each track is presumably OK (Spoon is not yet sued[*]), a "checksum" for each bit can identify the entire content and is not OK.

[*] Oh well, neither are the lyrics databases AFAIK, but I cannot imagine why they, technically, should be considered legal. IANAL, though.

[EliC]

Re: kick ass idea to make you super famous

I don't believe lyrics databases are legal, unless they pay fees.

I have asked/requested a more plug-in friendly interface. Who know what types of addons could be added to the ripper that spoon does not have the time to develop or want to for any various reasons.

It would be fantastic if someone could host a Reed-Solomon ECC database for correcting rips on the fly. I know it will probably never happen, but it would be the ultimate (when combined with AR2) in accurate ripping.

[Spoon]

Re: kick ass idea to make you super famous

There is no way they could sue for a small single checksum.

Lyrics databases are not legal, the song lyrics are copyrighted. Track listings cannot be copyrighted as they describe the contents (in the same way a telephone directory cannot be copyrighted), album art is a grey area.