I have some datasets that date back to 1965 that I maintain and are still used and updated.
Our Psychology Department does what they call longitudinal studies. They bring in a kid, age 2 or so, and give him/her some tests or make observations. They they track the kids and bring them back at ages like 5, 8, 12, 15, 18, 25, 30, 40, 50, 60, 70, etc and retest them Then they analyze the data and discover things like if you suck your thumb at age 2 you're 29% more likely to become a lawyer. OK I made that one up, but you get the idea.
The data was originally keyed in on punch cards submitted to an IBM mainframe running SAS for analysis. My predecessor transfered the data to 9 track tapes which were refreshed every few years. When our data center got rid of the mainframe and the last 9 track tape drive, we acquired the drive (a Kennedy) and bought a special interface card that allowed it to connect to an IBM PC ISA slot. The company that made the card had MSDOS software to control the tape drive and do EBCDIC to ASCII conversion. I put all the data onto 5 1/4 inch floppy disks with redundant copies. Later I considered moving it to CDRW, but read about limited lifetime of CDRW disks and decided not to. Instead I set up a Sun Sparcstation 5 file server and moved the data via ftp to spinning disk that was regularly monitored and backed up to QIC 40 tapes. I've upgraded and replaced that server 5 times, transferring the data via SCSI, ftp, and rsync. It's still online and will be after I retire.
Data format has not been too much of a problem. It's plain text files, 7 bit ASCII, fixed column width. SAS still works with those. The hard part has been the metadata. Columns of numbers don't mean much if you don't know the column titles. That's was kept at first on paper, handed down to each investigator that took over the project. Some papers were lost so I persuaded them to write the schema in plain text files saved with the data.
Slashdot Mirror
I have some datasets that date back to 1965 that I maintain and are still used and updated. Our Psychology Department does what they call longitudinal studies. They bring in a kid, age 2 or so, and give him/her some tests or make observations. They they track the kids and bring them back at ages like 5, 8, 12, 15, 18, 25, 30, 40, 50, 60, 70, etc and retest them Then they analyze the data and discover things like if you suck your thumb at age 2 you're 29% more likely to become a lawyer. OK I made that one up, but you get the idea. The data was originally keyed in on punch cards submitted to an IBM mainframe running SAS for analysis. My predecessor transfered the data to 9 track tapes which were refreshed every few years. When our data center got rid of the mainframe and the last 9 track tape drive, we acquired the drive (a Kennedy) and bought a special interface card that allowed it to connect to an IBM PC ISA slot. The company that made the card had MSDOS software to control the tape drive and do EBCDIC to ASCII conversion. I put all the data onto 5 1/4 inch floppy disks with redundant copies. Later I considered moving it to CDRW, but read about limited lifetime of CDRW disks and decided not to. Instead I set up a Sun Sparcstation 5 file server and moved the data via ftp to spinning disk that was regularly monitored and backed up to QIC 40 tapes. I've upgraded and replaced that server 5 times, transferring the data via SCSI, ftp, and rsync. It's still online and will be after I retire. Data format has not been too much of a problem. It's plain text files, 7 bit ASCII, fixed column width. SAS still works with those. The hard part has been the metadata. Columns of numbers don't mean much if you don't know the column titles. That's was kept at first on paper, handed down to each investigator that took over the project. Some papers were lost so I persuaded them to write the schema in plain text files saved with the data.
What do you get when you combine Beryllium and Argon ions? Be+Ar- A polar bear.