Yes, Allison, everything has a finite, if small, failure rate. I don't have
stat's on the failure probability of properly stored EPROMs in a plastic
package, but the failure rate of FD's is on the order of a part in 10^9 or
so. The random failure rate of a SCSI interface has been claimed to be a
part in 10^15, and I'd bet the random failure rate of a plastic packaged
EPROM is on the order a part in 10^20. Those are favorable odds, and, in
fact, I'd find the 10^15 quite acceptable for archives. Remember, we're not
looking at a random soft failure with FD's, it's the ultimate failure that
we're concerned with, since, once the archival copy is gone, the data on it
is lost forever. I'd be inclined to use a CD under such risk, wouldn't you?
It's an archive, after all, not a utility copy lying on the floor behind
your desk.
I'm not preaching that an OTP EPROM is the perfect solution, but, it takes
an hour or two to generate the hardware/software to read/write an EPROM.
If, OTOH, I had all the parts in perfect condition, in a kit, and a
well-lighted workstation with all the necessary tools, I still doubt I could
assemble an FDD, especially one of the little ones we use today in that
little time. They're available almost anywhere for $5 (used) to $20 (brand
new), but two years after they're replaced with something better, I doubt
you'll find one new drive for sale anywhere.
It's for that reason that I think all this effort put into magnetic storage
for an archive that will, hopefully outlive its creators, is a mite unwise.
Dick
----- Original Message -----
From: allisonp <allisonp_at_world.std.com>
To: <classiccmp_at_classiccmp.org>
Sent: Tuesday, May 30, 2000 6:28 PM
Subject: Re: Defining Disk Image Dump Standard
> >If you're serious about creating an archive. It needs to be permanent,
so
> >it's essentially requisite that the media be write-once. I'd say you're
> >better off with an OTP EPROM. Hardware to create them is dirt-simple to
> >create, eraseable/rewritable equivalents are readily available, and even
if
> >the OTP's (very inexpensive, by the way) become scarce, there will still
be
> >rewritables available which can be write protected by removing the VPP or
> >program pin. Use those and you'll have a real archive. What's more,
there
> >are no mechanical components, nothing to rust, become misaligned, or wear
> >out.
>
>
> Your kidding? Right? Eproms, have a finite random failure rate and while
> better
> in some ways over time you still run the risk of total failure due to:
>
> Environment, humidity increases failure rate.
> Temperature
> Time
> Electrical stress.
>
> This does not include ESD and circuit mishandling. it still assumes
> compatable technology (try reading a ECL prom using TTL).
>
> Add to this the great number of devices needed to contain said archive
> your risking the boat in exactly the same way as CDrom or on shorter
> time spans magnetic media.
>
> All of this is seperate from the format that will assure recovery of the
> data.
>
> Myself I'd rather risk even floppies with their known weakness and use
> redundant recording and added error detection/correction. Even when
> applied to CDroms this is more viable.
>
> Most of all it matters not what you do, what you do it on! It does matter
> that sufficient data on what was done is available along with the archive
> to reconstruct not only the data but the systems that archived it (or can
> recover it!!!). Books work because they can be preserved and copied
> if they start to decompose and we teach the languges needed to read
> them.
>
> Reminds me of an old theological arguement of how many angles can
> dance on the head of a pin. Untill we have music, a pin and angles
> it's simply an exercise with a meaningless outcome to all but the
faithful.
>
> Allison
>
Received on Wed May 31 2000 - 00:25:43 BST