Data Archival (OT Long)

From: Jerome Fine <jhfine_at_idirect.com>
Date: Fri Dec 8 19:42:34 2000

>McFadden, Mike wrote:

> [Snip]
> go. Of course their true interest is the next quarters profit margin.

Jerome Fine replies:

One word - GREED! I have rarely seen a company that believes
that their first responsibility is to meet the needs of the customer
as best they can. What is not realized is that over the long term, that
is also best for the vendor. The problem is that share prices over the
short run are more important.

The other problem is that no matter what the size of the online storage,
a backup MUST still be kept. I suppose that tape is OK for this
since it is only a matter of days, not even weeks.

> I am from the old school and like off line backups. Now the question is
> what kind of off line backup. I have old thermafax medical reports that are
> 35 years old and are basically unreadable. Film at least will survive if it
> is stored properly. I know we have 8" floppy data of CT images from the
> late 70's and probably 1600 bpi 9-track tapes. I don't even try to read
> them.

I still have some 8" floppy data as well, although only from the 1980s.
But, I no longer regard any tapes that are more than 5 years old as being
a reasonable media. Every tape more than 2 years old should be written
again to keep a backup current. Of course, I have some tapes that are
more than 10 years old, but I would never rely on them even if I had
20 copies unless I had just read at least 10 of the copies successfully
within the past 6 months.

> Everybody is converting their 16mm cardiac angiography movies from film to
> CD's.

No evidence, but CDs seem like the best media for the next 30 years.
Key point is that it is already machine readable - that aspect is so important!

> I think I need to develop a film based technology. There probably is some
> sort of microfilm already around.

I agree that microfilm is OK for some purposes, but each situation must
be looked at individually. I hear that NASA has barns full of tapes
which can no longer be read. If they could have been transferred to
CD before that happened, the data could have been saved. Likely
CD media and low cost did not arrive in time. The window of
opportunity was probably never there to start with - likely by more
than 10 years and maybe even 20 years of a mismatch before the
first tapes started to go bad.

> If you want real long term storage I have heard that there are people
> working on 10,000 year storage methods, etched iridium plates. For some
> reason I seem to remember that the genealogy people in Salt Lake City, Utah
> are working with this.

This is the real reason that I am sending a reply. If plans are being made to
save data for at least 10,000 years, then I suspect that it is also more
than reasonable to ask - no demand - that operating systems are able to
handle dates far into the future as opposed to just the next few years
in many cases. I am in the process of completing Y2K patches for an
old hobby version of V5.03 of the RT-11 Operating System. Fortunately,
it achieved hobby status and it is available for download. But I do not
think that stopping at year 2099 is reasonable at this point. The patches
should be able to carry the code to at least year 9999. I tend to feel
that if storage methods are being looked at which can last for 10,000 years,
then perhaps the almost trivial extra effort should also be made to allow
all code to handle up to the year 9999 at this point so long as any effort
is being made to extend the date window for a program in the first place.

Many people have told me that I am crazy to even consider this
possibility, let alone do the work, but why save data for 10,000 years?
Does anyone have any advice as to how to handle the problem that
the calendar can't be predicted exactly so far into the future?

Sincerely yours,

Jerome Fine
Received on Fri Dec 08 2000 - 19:42:34 GMT