Waaay off topic (was Re: Which scanners REALLY provide 36 bit output? HP?)

From: Nick Lamb (njl98r@ecs.soton.ac.uk)
Date: Sat Dec 16 2000 - 20:47:11 PST

  • Next message: Le Loarer Loic: "Re: Pb with listing devices when no root"

    Can those who are interested continue this off-list perhaps? I'm not sure
    we're achieving anything constructive, but if we are it's not SANE
    business any more.

    Steve Underwood continued:

    [Steve, it is conventional among those of us who don't expect our words
    to be read on parchment in 2350 to use 78 or less characters per line]

    > > But there are great advantages as well:
    > > 1) Ease of copying.
    > > 2) No loss of signal from copy to copy. Once a pixel is digitised into
    > > a "5" that 5 remains a 5 no matter which generation copy it is. And a
    > > weak five will copy to a strong five.
    > This is a weak argument in a world without perfect error correction. I
    > have _never_
    > seen this work out. I can make a thousand perfect copies of a disk
    > tomorrow, but the
    > same thing doesn't work out over time. There will always be uncorrectable
    > errors
    > over time, and often the media fails totally.

    Because your argument seems to be based purely on anecdotal evidence I'm
    fairly sure the statistical implications of what you're saying haven't
    sunk in yet.

    All digitial media copies are equivalent, and can be automatically and
    cheaply checked for differences at any specific level of statistical
    certainly / cost trade off. Don't put ANY of your eggs in one basket.

    So I can keep my documents on mirrored disks in New Zealand, AND on
    mirrored disks in Surrey. If a disk fails, I can copy a non-failed
    disk to a new disk, repeat. If a country erupts into war, I can build
    a new archive in a neutral country, transfer billions of documents in
    a briefcase-sized box, and lose nothing.

    I might lose all of it, all four disks on two continents failing
    irreversibly (we'll ignore the fact that most "irreversible" crashes
    can be recovered with an electron microscope, a commercially available
    service) but I can always reduce the chance of that by adding more copies
    if my only concern is future availability.

    Disk failure is not only rare, but predictable (bath tub curve) and it
    is flagged (will your paper warn you if a page fades due to a local
    hot spot?) by on-line systems very quickly. When the orange light comes
    on, you put in another disk, and throw away the disk with the orange
    light. Cheap and easy.

    Meanwhile, every archive of "real" documents is different, and checking
    for "significant" differences (page or line missing) even in text is
    extremely difficult, costs rapidly prohibit acceptable accuracy.

    Any such archive will be quickly destroyed by war or natural disaster
    (how much was lost during WWII?) and is difficult to retrieve under
    time pressure. Sheer volume is not on your side in an emergency. Making
    a copy archive of "real" documents is lossy and much more expensive
    than for digital documents.


    Source code, list archive, and docs: http://www.mostang.com/sane/
    To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com

    This archive was generated by hypermail 2b29 : Sat Dec 16 2000 - 20:43:29 PST