There are some quasi-plausible arguments for vinyl, but arguing that durability is one of them is hilariously myopic.
Gracenote had 3 598 785 CDs in their database in 2005 [1], so I'll take that as a lower bound on the number of CDs ever made.
I don't know what the largest flash drive that will fit in a jeans pocket is, but the largest I was able to find is 2 TB, but SanDisk showed a 4 TB prototype at CES 2019, so I'll go with that.
Assuming lossless compression that averages reducing the size of a song by 75%, a 4 TB flash drive could hold 1 662 149.1 minutes of music.
Unless the average CD is under 27.7 seconds long, that's not even enough to hold all the CDs known to Gracenote in 2005.
Anyone want to do the math if we allow lossy compression to squeeze more in (although I wouldn't consider it a backup of the CDs if lossy)?
https://www.wolframalpha.com/input/?i=177.5+cm%5E3%2F%28%28%...
That'll fit in my pocket, so that's 240 million megabit, 160 million seconds, or 2.7 million minutes, about 1 million songs. OK not every song ever, but a good portion, and possibly every song that's been in the charts.
I did see one claim that there had been as many as 100 million songs ever recorded, which would mean the density won't be there fore another 15 years or so.
An average person can expect to live upto about 2.5 billion seconds, which requires (for stereo 44.1khz 16 bit), about 220TB - 2.7TB a year. Given the growth of storage, you could record everything you hear from now to forever, keeping it in your pocket.
E: Specifically I'm talking about the P & Q channels, but this can also include R through W which while technically unused, still contain data in some form.
Most audio CD extractors do not provide all of this channel data, as only the digital audio is extracted. When a CD is re-burned, the subchannel data is regenerated, but not a bit-for-bit copy of the original. Sounds the same (usually), but sometimes there is metadata, easter-eggs, and what-not that is omitted. Additionally, some copy protection relied on adding corrupted subchannel data which low end CD players would skip, but high end extractors would not. Having the original subchannel data can be key and necessary to determine whether a sector was infact incorrectly extracted or not.
more information here
If all we’ve got is opinions, let’s go with mine:
None of the CDs I’ve ever bought still works. I’m probably not as careful with my things as you are, so there’s that.
Unfortunately, due to manufacturing defects there are discs that will or have oxidized, so the durability of CDs (or other optical media) is not always as reliable as advertised. But as the sibling post to yours points out, CDs provide error correction and along with oversampling provides means of creating effectively bit perfect copies that can be as durable as the current state of the art (redundant, solid state, geographically diverse, etc.)