Disasters and Recovery

From: Eric Smith <eric_at_brouhaha.com>
Date: Sun Jan 17 20:41:47 1999

Sam wrote:
> I would hope that in 100 years re-aligning a late-1990's hard drive would
> be a trivial task.

Doubtful. It's a hard task now, and it will only get harder, since the
necessary technical documentation (which is proprietary to the drive
manufacturer) will be long gone.

Personally I think that the best bet for 100 year data retention of large
amounts of data (such that punched tape or cards are impractical) is
CDR media kept in the dark under proper temperature and humidity control.
And make multiple copies on different brands of media.

Note that there are three important considerations to long-term data storage:

1) The media has to survive in a sufficiently good condition that it is
    possible to recover the data from it.

2) A suitable mechanism for the data recovery has to exist, or be
    constructed. Sufficient technical specifications of the media format
    has to be preserved to allow the preservation or construction of the
    mechanism.

3) Specifications for the data format must also be preserved. It does
    very little good to be able to recover raw binary data from the disc
    if you then are unable to interpret the data. For instance, a
    comma-delimited ASCII text file will probably be easier to read
    100 years from now than a Microsoft Excel '97 file. But carrying
    this idea one step further, there still could be questions about what
    the data represents. This is perhaps the most difficult problem.

The second consideration may actually require the use of media that is not
optimal for the first consideration. For instance, other forms of optical
WORM discs might have better longevity than CDR, but the technical
specifications are not as readily available.

For instance, people have suggested that in 100 years no one will be able to
read CDs. That may be true, but the CD-Audio format was specifically designed
to be easy to read, without even requiring a microprocessor (i.e., the P
subcode). As far as I know, no vendor ever sold a CD player that was not
microprocessor based, but the principle remains. And the CD-ROM format is a
fairly simple elaboration of the audio format. Technical documentation on the
CD-Audio and CD-ROM formats are so widely available now, that it is reasonable
to expect that a determined researcher 100 years from now should be able to
turn up copies.

Building a CD-Audio player without using only generic components (both
electrical and optical) with no parts specificially designed for CD players
could probably be accomplished in six months to a year by a small group of
researchers or graduate students. Expanding it into a CD-ROM drive would
not take too much additional effort.

Eric
Received on Sun Jan 17 1999 - 20:41:47 GMT