Sweet Spot for Error Recovery

[EliC]

With the batch ripper getting ready to go live soon it would be nice if the secure ripper could be set to find a sweet spot for secure ripping. In my mind this is:

1)accurately identify ALL errors
2)estimate the probability of recovery
-if reasonable:attempt to recover using current method
-if recovery fails after x sec quite and mark as error
-if unlikely: quite the rip and mark as error

The goal is to rip tracks that can be reasonably ripped and skip those that would take forever.

[bhoar]

Re: Sweet Spot for Error Recovery

With the batch ripper getting ready to go live soon it would be nice if the secure ripper could be set to find a sweet spot for secure ripping. In my mind this is:

1)accurately identify ALL errors
2)estimate the probability of recovery
-if reasonable:attempt to recover using current method
-if recovery fails after x sec quite and mark as error
-if unlikely: quite the rip and mark as error

The goal is to rip tracks that can be reasonably ripped and skip those that would take forever.

Some similar ideas that have been bouncing around in my head (that might be possible to mix and match):

1. Just as Eli said: a threshold upon which it will discard a track and move on to the next track. Perhaps it could be linked to a basic timeout (e.g. x=10 minutes per track...that wouldn't be good for long tracks...). Hmm... Better yet a weighted timeout (e.g. x=10 minutes per minute of track time?) which would be equivalent to a low-water ripping speed (e.g. 0.10x) over some time interval. Or just directly set a low-water ripping speed and time interval that seems to be a useless waste of your drive lifetime...

Essentially, some sort of substitute for the OCD operator who hovers and often clicks skip track and cancel disc.

2. A sampling style two pass approach. An initial, no codec, sampling pass (to get general track health, e.g. reads 1s for every 30 seconds of audio...whatever seems like a reasonable sampling strategy that takes into consideration seeking/settling/caching behavior). A second ripping pass that rips the tracks in order of track health. That way, if you *do* cancel the disc (manually or due to a timeout) due to a track #2 that is causing it to take 20 minutes, you aren't cancelling out tracks 3-10 which are good.

3. A two pass approach (note: i recognize this poses performance issues with the batcher->ripper->compressor chaining)?

A low-threshold pass that auto-skips troublesome tracks.

A high-threshold pass that handles those tracks that were autoskipped.

This is almost parallel to the way that the secure ripper does rips (pseudo-burst w/ validation first, more intensive corrective rips follow), but instead of trying all strategies in serial on a single track it runs the tracks through several ripping strategies. E.g. the bursty ripping first to get the most easy to grab (validated) data early in the rip time.

I can see how this would also present memory issues (if you kept the initial track rip in memory when skipping) or performance downsides (requires even more rerips of failing tracks). So, #3 is probably not the best idea out there.

-brendan

[EliC]

Re: Sweet Spot for Error Recovery

some sort of automated feedback from the ripper to illustrate may help give data to guide this.

[EliC]

Re: Sweet Spot for Error Recovery

I think clearly one area that could be set is the number of frames that require a re-rip. If it is >x, mark as error. It would be interesting if dBpoweramp could get feedback from the ripper in the wild and see, on average, when people are using the secure ripper, if >x frames need to be re-ripped how often is a secure rip successful?

[EliC]

Re: Sweet Spot for Error Recovery

I would also be interested to know how many re-reads on the frames on average before you can be relatively sure that there will not be a good read. IE, after 5 re-reads of a frame what is the chance. The current default settings maximize the chance of getting the correct read. For me, I want to know if I can get a correct read in a reasonable time and if its unlikely I want to skip the track.