On Fri, Apr 17, 2009 at 10:22:06AM -0400, [email protected] wrote:
>
> It depends.
Indeed it does - with failure (as in all cases) being the dominant
option. Second Law of Thermodynamics, you see. :)
> Ever since the days of popular music on magnetic tape it has been
> understood by many people that magnetic recordings fade with time. The
> floppy drive in my laptop still works well but most of my old disks have
> read errors.
Actually, mechanical data loss is a relatively controllable factor - and
is usually not the key problem, although it certainly can be. Most of
the problems in this area tend to lie in the area of lost metadata -
i.e., data structure/format, media type, knowledge of what the data
*is* or means, etc.
As an example, if you have say, a magnetic tape that you need to read,
here are the problems you need to solve:
1) You hope and pray that neither the tape itself not the ferrous medium
on it has deteriorated. Nothing you can do if it has, so you go on to
the next step.
2) You need to find the hardware to read it. If you can't - either it
doesn't exist any longer, or your research on the topic comes up with no
answers - you'll need to design the equipment. That is, after you've
figured out the recording/playback method: necessary tape speed,
recording level, type and level of bias, etc.
3) Assuming you've managed the above two hurdles, you now have two piles
of bytes. HOW DO YOU KNOW WHAT THEY REPRESENT??? They could be
a) A headerless stream of data - e.g., everything from 1h42m12s to
2h03m06s of a space noise recording;
b) Another type of headerless stream of data - the last 22% of file
A, all of file B, and 92% of file C (the index data is stored on a
tape that you don't have);
c) A regularly-delimited data stream, where every 112th-120th byte
represents an error correction sum for the preceding 111 bytes;
d) A descriptive header containing indexes for the data to follow,
with an ECC byte after every 8 bytes... but since you have no idea
of the format of that header, it's *still* just a pile of bytes;
e) Option d), but compressed using an algorithm that was only ever
used by the company that produced that backup software;
e) ...an infinite variety of other options.
At this point, unless you can track down the metadata (i.e., information
about how that data is structured), you're well and truly stuck. Even if
you have NSA's big code-cracking computers, you have nothing to start
from in order to form your guesses. Game, set, and match - with a total
loss for the person doing the guessing.
> When I got my first CD burner (A rugged Sony that we still use. It is the
> only one we have that will read discs with paper labels. Paper labels make
> the discs warp when they get warm causing them to jamb in other drives.) I
> went through all my 3 1/2" floppies and copied everything I wanted to save
> onto cd's, then later combined them onto DVDs.
>
> >From my investigations, the best optical media seems to be Taiyo-Yuden
> DVD+R media (supermediastore.com). DVD-R media is said to be a kludge
> cobbled up from CD recording techniques, whereas DVD+R was made entirely
> new for DVD recording.
I recall reading something like that myself. There's also "archival
quality" optical media available; unfortunately, only a few drives can
take full advantage of it. NIST and Johns Hopkins have been working on
coming up with a certification for 200-year media, but I haven't heard
whether they've managed it yet or not.
> I have also read that optical (CD and DVD) disks routinely have a large
> number of errors during apparently normal reads, but the high quality of
> error-correcting software now in use makes us unaware of this.
Actually, that's true of most types of storage media...
> Apparently I should transfer my optical archives now onto thumb drives or
> similar flash memory.
...flash memory included. Error correction (ECC) circuitry is a standard
part of every hard drive controller, for example.
> However, in 1860,
> 17 years before Edison's machine, a Frenchman smoked a piece of paper, then
> scratched a line on it with a needle connected to a diaphragm thus making a
> recording. This recording was not meant to be played but simply observed.
> This artifact was recently scanned by laser and the recording, someone
> singing an excerpt from "Au clair de la lune", was recovered.
> www.firstsounds.org/sounds/
Please note the (implied) cost _and_ simplicity of reconstruction: the
exact method of recording was known, the format of the data being
recorded was both known _and_ amenable to a high degree of data loss (if
you lose 50% of a piece of music, it should still be recognizable.)
Digital data, by contrast, is not very tolerant of loss, has a nearly
infinite number of formats, and is not easily recoverable.
> It would thus appear that where there is a will there is a way to recover
> obsolete data.
In extraordinarily simple situations in which all possible factors are
in favor, yes. Anything less than that, and the chances rapidly start
circling the drain.
--
* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://LinuxGazette.NET *
_______________________________________________
Liveaboard mailing list
[email protected]
To adjust your membership settings over the web
http://www.liveaboardnow.org/mailman/listinfo/liveaboard
To subscribe send an email to [email protected]
To unsubscribe send an email to [email protected]
The archives are at http://www.liveaboardnow.org/pipermail/liveaboard/
To search the archives http://www.mail-archive.com/[email protected]
The Mailman Users Guide can be found here
http://www.gnu.org/software/mailman/mailman-member/index.html
