Hacker News Reader

Comment by andix | Hacker News Reader

For all the younger people I think it's important to know, that before Windows 95 the concept of "shutting down" a computer was mostly unknown.

DOS never had an option to shut down, you just saved your work and turned off the computer. Also with Windows 3.x it was normal to just switch the PC off, without any prior shut down process.

Back in the days most software was synchronous, there were no background processes, no write cache (and read cache was optional). Once the "saving document" message was gone, it was written to disk. With floppys and early hard drives it was easy to hear anyway, when the noise stopped it was written to disk.

View on HN

> it was easy to hear anyway

It has been very frustrating to watch designers ramping up reliance on the side channels by piling functionality into a single button at the same time as the side channels were disappearing and software latency was telescoping so that in present year I frequently have to reverse engineer morse code power button bullshit with 20 seconds of lag in the tight loop before the first audible/visible cue comes through. Is a short press 0.5s-1.5s? Or is that a long press? Is 0-0.5s ignored as debounce or is it a short press? Does a long press suffice for shutdown, or do you need a really long press to trigger hardware shutdown because the software shutdown is unreliable? Is 5 seconds long or really long? AAaaaaaaaaaa

Even in the DOS days, you still had to park the drive heads before shutting down. DOS even had a PARK command for that.

True, that was a good idea with some early hard drives, but only required if the PC was moved. In that period hard drives were a luxury, many pcs just had two floppy drives. One for the operating system and applications, the second one for data.

Only on early hard drives that used stepper motor actuators.

...but most people got their systems from their job in those days and so wouldn't get told this little tidbit for obvious reasons; You Must Park Before Swithing Off. I did get unlucky somewhere begin 80s when my father brought home a luggable, I didn't park, he lugged it back and the hd was damaged.

Yes, the GP is a more recent advancement. Before that, you had to set the hardware for turning off.

There was a brief moment when the hardware was advanced enough to reach a safe configuration by itself, and the software was not advanced enough to need to synchronize state with everything.

But maybe there's a lesson there, and the software should be able to shut itself down in a couple of seconds in the background, just like the hardware did.

> DOS never had an option to shut down, you just saved your work and turned off the computer.

I miss this simplicity. With journaled filesystems, it should be pretty safe to do this.

Few things bother me as much as a computer that takes more than 1500ms to shut down.

Many modern embedded operating systems support this (including some Linux distributions and Windows IoT). It’s possible, just not very practical, most wouldn’t consider the necessary trade offs worth it.

If you boot Windows from a RAMdrive you can do this.

Paraphrasing: If you boot <OS> in a way that it will never save data, you can do this.

AIUI there can still be write caches on SSDs etc

I'm not even aware of anything stating it was safe to kill power inside Win 3.xx

Exiting Program Manager before flipping a power switch was always normal operating procedure.

If you killed power in Windows with no regrets you've lived a dangerous life!

> If you killed power in Windows with no regrets you've lived a dangerous life

Yes. You better were sure that your harddrive light was not blinking, before pressing the power switch.

Indeed. DOS had primitive disk caching, but it flushed all caches before returning to the command prompt, for continuity with the time when it had no caching at all. So if you saw C:\>, you could safely turn the computer off. You wanted to exit Windows before cutting the power.

Another wrinkle to this is that many computers' power switches were physical switches on the power supply that directly closed or opened the circuit supplying power to the computer. There was no way for the computer to switch itself off like in today's ACPI-mediated world, so the user had to be told when to turn the computer off.

Just goes to show how backward and primitive the PC was until well into the 90s. Macs had been shutting themselves off, and ejecting their floppy drives, since 1984.

Shutting down Windows 3 (mid-range) corrupted a project I had been working for months on. One still had to take care.

Sure, if you turned it off too early the files were corrupted. The safe thing was to exit Windows 3.x to the DOS prompt and only turn off the computer then. But it wasn't really required.

Everyone I knew would fallback to DOS before throwing the big red power switch else risk suffering a day of work or worse being corrupt tomorrow morning.

Edit: the windows 3.1 manual "recommends" exiting windows before turning off the PC. It doesn't say it's "required". And with 386 mode, windows actually did do disk IO, including caching.

Probably a lot of people did exit windows 3.x before powering off the computer.

I actually don't think it was required. Windows 3.x was just a shell on top of DOS and didn't do any IO caching. Saving files or closing all applications should've been safe enough.

> that before Windows 95 the concept of "shutting down" a computer was mostly unknown.

I think you mean a "personal computer". Apollo DomainOS, at least, had a shutdown command and i presume that also VMS had one.

Reply on news.ycombinator.com