"The problem is, magnetic tapes are regarded by storage and archivist professionals as being a stable, reliable, and safe medium for long-term data storage. Just because it’s a 70 year old medium does not’t mean those records needed a massive overhaul to digital, that it will save any money in the long term, or that the new storage method is better."
This is a false statement.
In any listing of storage and archivist professionals... I would be considered as one of the top experts in the field.
I have served on the IEEE Data Migration Standards Committee as "chairperson". I stored the nations climate data from 1993 to 2000, I have been in the forefront of Data Storage for supercomputers for 40 years.
Tape is a digital format. Worse, tape does not last that long, even in climate controlled conditions. The mylar backing deteriorates over time. the iron oxide actually falls off the tape. The read and write heads tend to scrape the iron oxide off because this is a contact media. The tape also can "stick together" so one must spin the tape from source reel to take up reel and back to source reel every year or 2, and not on a tape drive with read and write heads in contact with the tape. And have you tried to buy a decent dual vacuum reel to reel tape drive recently? I think they are only in museums.
And while it is easier to corrupt ( silently ) spinning disk, solid state or flash devices or even storage class memory, it is very possible to corrupt data on tape without leaving evidence as well. Truly safe media for long term storage would be write once media such as WORM optical... or anything that physically etches the media ( electron beams, lasers, etc ).
Then you get into the field of provenance, i.e. the ability to document the data you are storing and the provability that what you stored is still intact and unaltered. I would refer people to one Margo Seltzer, a professor at Harvard ( maybe retired now, idk ) who I sponsored when I was at NASA ( NASA has a large stake in keeping digital data around for decades, considering the cost to the taxpayer of obtaining that data ).
Last thing for data archivists to face is basically the hieroglyph problem. This is to say that data written by software and hardware in combination with each other will be difficult to decipher decades later when, even if the media survives, the hardware and software has NOT survived in working condition. A problem I faced when I joined NASA and was given a large pile of 7track 256 BPI reel to reel tapes from the Mariner Missions to save and restore for researchers to discover changes on Mars from the 1970s to present. This only real solution is to copy the data forward on a predictable technological schedule, about every 4 to 5 years as computer technology changes. This includes an effort to move the software forward as well, one can imagine moving from RPG to COBOL to Pascal to C to python ( with possible intermediate stops along the way ). Moving the data dictionaries forward as well. And finally, and worse, moving representations of numbers and characters forward... from 1's complement to 2's complement, from 36 bit IBM 1960s machines to 60 bit CDC to IEEE 64 bit floating point and from EBCDIC to 6/12 ASCII or 9 bit ASCII to standard ASCII. All without losing either precision or meaning.
My objection to DOGE is that they have NO experts doing this and they are trying to do it slap-dash with no understanding of the code that wrote the data ( COBOL ) and what the data actually means. The age of 150 years for example... probably a selected fake birth date of someone for which the records of date of birth are unclear and so a default date was selected.
In UNIX systems, time is a very tricky problem... as time "0" actually means a specific date in 1970. The entire year 2000 problem ( in many systems we only recorded the year as 2 digits... oops ). There is another time issue coming up, 03:14:07 UTC on January 19, 2038 will be when all unix systems using 32 bit time will also "wrap around to 0". There are many flavors of unix ( all Linux systems, all Apple systems, many embedded computers ) where this might cause an issue.
= new reply since forum marked as read
