You're both saying approximately the same thing. The media has gone bad because it lost one or more charged bits.
No, I feel we do not. First the assumption that data will always be a 0 or 1 is incorrect IMO, an over simplification. The media simply does not store zeros and ones. And it does not have to go bad for bits to flip.
Depends on how you define 'data'. A user is not interested in storing the charge level. The digital
data is 0s and 1s.
I can argue a user isn't interested in zeros and ones either. He stores a document, or a photo so I do not think that's a strong argument. And yes a computer handles all that data as zeros and ones. But once written to storage it is no longer zeros and ones.
Zeros and ones is an abstraction. The meaningful
data being recorded is a sequence of 0s and 1s.
You're twisting this. All I was saying user is as less interested in charge levels as he is in zeros and ones. To the user everything that happens after he clicks the save button s a black box. He does not care about charge level, nor zeros and ones. To put is simply, I don't see what a 'user' has to do with this.
My point is that storage media store analog information, not digital data.
We intend to store zeros and ones, but we don't.
We intend and we do. We do store zeros and ones recorded/encoded as changes in the physical state of the media. We then need to follow a certain protocol to read them back.
A battery will leak charge over time, not because it's bad but because that's just the way it is. The outcome may be undesired because we have no power or in case of NAND memory the bit flips, but it's not due to bad media. We discussed 'bit rot', bit rot can occur on perfectly healthy media. It occurs because information stored as not as binary as we intend it to be. It's truly decay, it's not a one that drops all of a sudden, it gradually gets closer to a threshold until it's no longer decoded as a 1 but a 0 instead or vice versa.
Yes. Which means there was a physical change in the corresponding medium. Or something happened in the controller (e.g. voltage became too low for whatever reason). Or there's a parasitic capacitance because of a manufacturing defect.
No, there wasn't a change in the medium, there was a change in information it contained.
In my book media is 'bad' when a sector becomes unusable. Which is not the case here.
It's 'bad' when it breaks the protocol: you encoded a 1 and it reads a 0. Whether it's recoverable and whether it's marked as 'bad' is another story.
It isn't. It's functional, like I said if the decay is detected early enough data can be refreshed, and this is possible because the media isn't actually bad.
You have no clue if it's actually functional or not, all you can do is to follow the reading/decoding protocol and verify the result against some redundancy data (parity bits, CRC etc.)
No, it's bad when media is unfit to store data is my point. If I write data and I can read it back then it's not bad, it's functional.
Media is bad, at least that's my definition, it's all a matter of definition, when it loses it's capability to store data. A bad sector can not be read from or written to.
It's bad when it loses data (fails to read/decode after following the protocol). You write A, but read B.
That's not due to bad media, it's due to bad information. That's my whole point.
Whether or not the controller or operating system marks the sector (or the whole drive) as 'bad' is another story.
Which was exactly my point. It will only do so if it is unable to store new data. The information being corrupt is not a measure for the media being bad.
With similar errors, different controllers/systems may mark or not mark the faulty sector as 'bad' and they may employ different recovery protocols.
What matters for this conversation - 'bit rot' always happens due to unexpected changes in the medium.
No, it happens due to changes in analog information.
Point is, it is not possible to write plain zeros and ones to storage devices it's simply to 'fuzzy' for that.
Depends on the implementation (again in HDD 0s and 1s are encoded by means of flipping magnetic dipoles which map to {0, 1} set exactly.
No, that's not how this works. It's a simplistic layman interpretation.
The rest I ignored, it's same point over and over IMO.
Picture a grid of 8 x 8 buckets. It it allows us to store digital data, 64 bits. In your world the zero / one is represented by a bucket being full or empty if I correctly interpret what you're writing.
But PC storage does not work like that, it works more like: If bucket is > 50% full we interpret this as a one. If bucket is < 50% full it's a zero. So IOW we measure ranges, and if within certain range it's either a 0 or 1. So whether a bucket is 56% filled or 78% filled, we decode that as a 1. And whether a bucket is 0%, 23% filled or 48% we decode it as a 0. This is pretty much how NAND works. The information itself is not digital, where we set the threshold and decide what counts as 0 or 1 is what makes it digital. Magnetic storage also is interpretation of voltage spike due to flux reversal and thresholds we decide upon. The information itself is not digital, it's much more fluid.
We'd like to store zeros and ones, but we don't, simply because the physical world doesn't allow us. Instead we have to work with analog information the best we can.
Now without changing the media (the buckets) but due to external causes the analog information inside the buckets (amount of fluid) can change (rain, drought). In NAND this could for example translate to read/write disturbance, we change analog information in cells we do not intend to change. But even without those external factors and disturbances, the fluid slowly evaporates like a NAND cell will slowly leak electrons. It's stuff like this that results in what we tend to describe (I believe) as bit rot.
IOW, the media did not deteriorate, The analog information did. If we detect it using ECC for example we can even correct it on the fly but also update the information in the buckets
as long as the media itself isn't bad.
Over time indeed, due to for example corrosion buckets may lose some capacity to hold fluid, it simply leaks out.
That's bad media in my book. The media (buckets) lost it's ability to store information. Or a buckets gets punctured, like a hard drive that was dropped, heads scratching the surface as a result, scraping off the layer of magnetic material, that's bad media in my book and it's clearly not what's meant by bit rot.
--
Joep
, so simplified boot sector + FAT * (number of FATs) + cluster * (sectors/cluster).
