I have all my work (I'm a scientific programmer) back to the 60s backed up on mylar punch tape, from the 70s on eight inch floppies (Dec), the 80's on various mini-tap drives and the 90s various media storage including Iomega Zip drives. I would migrate my older code to the newer storage method except my mylar punch tape from the 60s which I would need a teletype to read into a more modern. By the time CDs were proven I had migrated all of my code to a duo-back of CDs, one keep at work and one keep at home (About fourty CDs for the date with a equal back up at a different location. I moved this stuff to DVDs and now have about 16 (each two copies one at a friends house) and all zipped on a five TB on line. I still have all the original media and devices boxes away. You'd be surprised when you might need a algorithm you coded in 1992 with the help of a russian speaking post docurate because Chebyschev's work was never translated from russian to english. I can lay my hands on the original K&R C code I wrote in two minutes and convert it to C#, Java, Pearl or whatever I'm using at the moment. I hate to have to repeat a development process.
I suspect not everone has a need for earlier development code or old data; as my EE friend says "if you haven't used it in the last two years, you don't need it anymore." While I think she needs to add the modifier probably won't need it.
While I am surprised how much I relie on old code (tight code still requires pointers and creae data access, especially if it needs to fit on a 2K rom) and I do so damn much windows net programming nowadays it helps to review what I did 30 years ago in FORTRAN or K&R C to get back to speed on old talents.