[OT] paper on Retro ?

From: Eric Smith <eric_at_brouhaha.com>
Date: Thu Oct 17 20:11:01 2002

> While we're on the subject, did anyone see the article "Data
> Extinction" in the October 2002 issue of Technology Review?

Yeah, but I was somewhat disappointed. The cover made it sound like they
would discuss *media* that would last a long time, but they were just
talking about maintaining the ability for software to handle the data.
That's obviously worthwhile, but it's also much easier to figure out:

1) Use industry-standard data formats, not proprietary formats
2) Use text-based formats where possible rather than binary formats 3)
Keep a specification for the format along with that data, ideally
    on the same physical medium
4) Keep the source code for the software that processes the data,
    ideally on the same physical medium. Ideally use software written
in a programming language that is widely used, not something
obscure.
5) Keep the tools used to build the binary of the software.
etc.

On the media side, though, AFAIK there is only one high-density
machine-readable media that can be expected with any confidence to last
much more than 20 years. That is CD-R with a gold-coated substrate
rather than silver or aluminum. Note that gold-tinted dye is NOT
sufficient. The main failure mechanism for CD-Rs is oxidation of
the reflective layer, and a gold layer won't do that, so then your limit
is based on other physical processes that occur even more slowly.
Kodak's accelerated aging tests indicated a typical life of several
hundred years, so they conservatively claimed one hundred. They used to
have a white paper on their web site.

However, almost everyone that made gold CD-R media (including Kodak) has
abandoned it. Mitsui appears to still offer it. Of course, it's more
expensive than the cheap stuff by a factor of two or more.

Kodak came up with a "marketing breakthrough" in offering discs with a
blend of silver and gold. They claim these to be better than silver
only discs, though not as good as gold-only. I am very dubious that
these mixed discs will actually hold up better than silver-only, since
there's not actually enough gold to be sufficiently reflective if all of
the silver oxidizes. But I don't really understand the chemistry of it,
so perhaps somehow the mix keeps the silver from oxidizing as quickly?

I speculate, but have *no* evidence, that gold CD-RW media might last
even longer than gold CD-R media. This is because it takes a higher
energy level to cause a phase change in CD-RW media.

If you need to archive data with any media with a limited lifetime
(which means any media at all), you have several potential problems.
You need to recopy the media before it fails, and you need to make sure
that you don't introduce any errors in replication. The longer the
lifetime of the original media, the less of a problem this will be.

When you do recopy the media, you'll probably want (or need) to copy it
onto a different, more modern medium.

One of the things people have worried about is how long CD-ROM drives
will exist. I believe that twenty-five years from now it will not be
difficult to find a working drive that can read CD-ROMs, but that in
fifty nears it will be somewhat difficult. However, one advantage of
the CD format is that it was designed to be so simple that a
microprocessor is not even necessary in a player or drive. This was a
serious design consideration in the late 1970s when Sony and Philips
were developing the CD for audio use, though by the time they introduced
the first actual products in 1983 the cost of a microprocessor was so
low as to make "dumb" CD players impractical.

The CD and CD-Audio specifications are widely available. You can't get
the official specs without paying some money, but they're in IEC 60908
(CD-Audio) and ISO/IEC 10149 (CD-ROM extension) which are *much* less
expensive than buying the Red Book and Yellow Bookfrom Philips.

I believe that building a crude but working CD-ROM drive from scratch
would be a reasonable project for a few grad students, so if it is really
the case that in 50 or 100 years, no new CD-ROM-compatible drives are
commercially available and no old ones still work, it won't be an
insurmountable challenge to read CD-Rs.

If you want a medium with even better longevity than that, I suggest
punched mylar tape. But the density is orders of magnitude worse than
for CD-R.
Received on Thu Oct 17 2002 - 20:11:01 BST