[semi-OT] QIC2 streaming cartridge tape lossage

[Michael ODonnell]

I finally got around to the task of re-archiving the data in my mongrel
collection of tapes (QIC2 format 1/4" Streaming Cartridge) that's been
nicely stored in its two copier-paper boxes since forever.  I figured
I could reduce the physical space required down to a handful of DVDs
while simultaneously moving to a format not associated with steam power.

Although I was in the habit of verifying readbility after creating
archive tapes I apparently waited too long to engage in this exercise
because it's been a very disappointing experience and I've only been
able to recover a portion of my data.  Luckily, I've not lost anything
too precious but on the (admittedly slim) chance that anybody else out
there is using this ancient medium I'd like to urge you to immediately
change over to, well, basically *anything* more reliable like, say,
smoke signals in a hurricane.

Most of my tapes were written between 1993 and 2001.  Brand names
include 3M, Wang, DEI, Carlisle, Maxell, Platinum, ICL and Verbatim
along with various unbranded distro tapes and such.  Regardless of
brand, I've seen such a poor yield that at this point I'm pleasantly
surprised when I'm able to fully retrieve data from any of them.

I claim that my kit isn't the problem - I have use of two Archive Viper
SCSI drives that in their day were top-notch, and most of these tapes
were written by the same model of (and in many instances the self-same)
drive which is, of course, not supposed to matter anyway.  Data path is
from external enclosure via high-grade SCSI cable, properly terminated,
using an Adaptec 2930B that's ID'd by the PCI subsystem as AIC-7860.
Kernel is 2.6.30-2-686.  General operability is seemingly confirmed by
the fact that some tapes can be read at will, time after time.

I'm seeing several types of mechanical failure related to the internal
drive belt loop inside the cartridge:

 - The belt has stretched and won't move the tape.  (workaround: use
   one from another cannibalized cartridge)

 - The belt breaks during operation.  (workaround: see above)

 - Slack tape fails to maintain proper contact with head and "retension"
   commands don't have desired effect.  (workaround: disassemble cartridge
   and retension by hand)

Fortunately, cartridges from most manufacturers are easy to open up for
emergency surgery, though their parts are usually not interchangeable.

I'm also seeing various medium failures where the bits just seem to
have fallen off the tape and are irretrievable despite heroic efforts by
the drive firmware to retry, reposition, re-align, rewind, recalibrate,
re-whatever ad nauseum.  It's obvious from the annoying "shoe-shining"
sound when the drive is having trouble.  Nice how the 9-track serpentine
layout makes it possible for a single medium flaw to trash the bit stream
in 9 different places at once...   >-/