And, in any case, I still have a paper here, ...
It was indeed the tradition that the full specification and full interface manuals (device-specific) for all disks were published. That tradition seems to have ended about a decade or two ago. The same is true for many other devices, such as CPU chips (look at the chip in the Raspberry Pi as an example of non-public documentation). We can moan about it, but it's a fact. And it is an effect of our litigious and dishonest society; I can fully understand manufacturers who keep their secrets, for their own protection.
I am talking about disks that definitely look different (because one is red, the other green, then purple, blue, etc.), but on the (very sparse) data that is published show no relevant technical differences.
Sorry, but they do have considerable technical differences. Those may not be published though. And if you go drinking beer with the actual people in the disk drive industry (which I do regularly), or you have access to the detailed technical information under NDA (used to do that too) you will hear about all these interesting factoids, and how disk drives are rather complex. Much more so than people can determine from the outside.
(SCSI vs. AT is a relevant difference, ...
Actually, rather on the contrary. In the 90s and early 2000s, many disk drive users used to think that SCSI and IDE disks were fundamentally the same mechanism (same spindle, same platters, same heads, same actuator), just with a different control board slapped on. That may have been true in the 80s, but it was definitely untrue in the late 90s and early 2000s. In those days, SCSI disks were built for and intended for enterprise use, which means 24x7, mounted in enclosures, controlled environments, often better seek times at the expense of more power consumption, and intended for reliability in the face of continuous operation and continuous seeking. That meant they were equipped with extra air internal filters (mostly to absorb lubricant sputtered by the spindle and actuator bearings), different spindle bearings and motors to survive 24x7 rotation, very different actuator and arm mechanics to handle continuous and faster seeks without heat problems, and extra CPU power (usually a second DSP) for servoing while doing IO, to prevent off-track writes while in the presence of vibration from dozens of other drives also madly seeking in the same enclosure. Consumer disks were sold with IDE = ATA interfaces, and were the opposite. They were also much cheaper. Some reckless small systems integrators were trying to build enterprise-class system using (cheap) IDE drives, which usually led to disaster. Uninformed people used to say that the disk vendors are financially raping the big enterprise customers, which is not just nonsense, but paranoid and dumb nonsense: in reality, the enterprise market is much more competitive and has razor-thin margins; if a large customer (back then EMC, HP, IBM or Sun, today Amazon, Facebook, Google or Microsoft) buys a few million drives, they squeeze the last penny out of the disk makers. Whereas the consumers who get their drives from Fry's or Newegg actually pay a huge premium for the convenience of getting individually wrapped disks.
Into this environment, where a lot of users used to think "IDE = cheap version of otherwise identical SCSI = expensive" came Erik and Dave, and wrote that paper to explain to people: no, the two drives are actually radically different. And the real difference is not the interface, but the stuff inside the box.
Now in the meantime, much has changed. Today, you can often get the exact same drive (mechanical assembly) with either a SAS or SATA interface. And often, the only difference between the two is that they populate a different connector (the SAS connector has a different polarizing tab and a few extra contacts), and load different interface firmware. Otherwise the drives may be totally identical. On the other hand, all the vendors sell drives that are actually build radically different, but you can no longer distinguish them by what interface is on them. And the differences have become even larger than what I explained above. Today, drives have much more firmware in them, and their performance characteristics are much more finally tuned to the intended usage. That begins with the way the data is laid out (zoning, cylinder versus platter layout), the sparing and re-location strategies, how multiple queued IOs are handled (optimized for low latency or high throughput). Today different drives of the same nominal RPM and nominal capacity will sometimes have different number of platters, different media, different heads, different lubricants on the platter, and the heads are adjusted for different fly height. Matter-of-fact, the really large customers (over 90% of all disks are sold to a handful customers) tend to special-order firmware versions that are optimized for their wishes: if Seagate or WD sell the model XYZ to IBM, EMC, Microsoft and Amazon, they will probably have different part numbers, customer-specific firmware, and different performance characteristics. Yes, I used to work for some of those disk customers, and occasionally I would get prototype "generic" firmware versions into my lab, and the disks performed differently. And I've been in meetings with people whose badges had words like LSI, Broadcom, Seagate, Hitachi, Toshiba, and WD, to discuss exactly these kinds of specific details.
So, as long as nobody comes up with a profound statement about what the technical differences actually are, I just assume there are no significant ones.
You are assuming wrong. Your paranoia is getting the better of you. My only advice is this: Either believe what the disk vendors say (and this particular example is very troubling, in that WD was caught being not truthful about these disks), or buy a few tenthousand disks and have your lawyers negotiate getting the full technical documentation. Simply assuming that everyone is out to get you is not going to be useful.
Any drive is capable running 24/7 if it can run 8 hours. It doesn't get tired and need to relax after 8 hours.
Patently false. A drive that is designed for 8 hours per day 5 days a week may not survive long if you don't let it cool down regularly, or if you run the spindle all the time (lubricant is a consumable). And that's just the relatively simple mechanical aspect. Now add the complexity of modern drives (fly height, platter lubricant, the fact that heads dive down for reading and writing), and drive-internal maintenance operations (garbage collection, validation, rearranging data) which depend on bursty workloads and sparse layouts, and usage patterns heavily interact with the design of the drive.
I only remember trade fairs at the end of the last century where manufacturers showed open drives in continuous seek.
Those were fun. It's even more fun to see a traditional "washing machine" drive (the 60MB top-loaders with exposed head control mechanism) seeking. I think the absolute maximum fun thing is to see a RAMAC in operation; and while I used to work nearly near them (there was one in the lobby of the building that contained my office), I've never actually seen one move. Here is a youtube film:
View: https://www.youtube.com/watch?v=oyWsdS1h-TM
, about halfway through you can see the arm moving (there was only one arm for all 25 platters).
That might be the only point where there actually is a difference between desktop and enterprise models.
Absolutely incorrect. Matter-of-fact, there are even significant differences between different commercial models.
(about laptop drives)
I got those to die in no time. They don't stand my usage patterns..
QED. Doctor, it hurts when I do that. Well, stop doing it.
I have no idea about the history of that drive, the smart values told something about 100 days operation and 400TB read. I have now found a suitable configuration, but there is no documentation at all about it, even worse, the specs say the drive does not support APM. (APM is something different than the idle timer; both are independent features.)