> I haven’t had anything since that could go so long between bounces.

Er. I've seen servers accidentally get into years of uptime from neglect; modern OSs are pretty stable.

that's a joke, right? I recently rebooted my Linux desktop after 2+ years of uptime because a random opengl app finally crashed kernel display driver.

And servers which do just one thing can keep going for years, until power fails or its time to decommission them.

> I haven’t had anything since that could go so long between bounces.

.... Really? Ignoring the "nothing should have that uptime due to security patches" best practice, almost everything should be able to do that barring power or hardware failure.

My OpenIndiana NAS has done >1 year uptime two or three times now. My OpnSense firewall did once (too much road warrioring for a period) but could do it easily again. Linux VMs have done that many times for me when I've been lazy on patching or too busy My ESXi hosts do that too. Thats just my basement homelab, nothing special.

I installed a cluster of Windows 2003 servers in a customer site a looong time ago that did over 1000 days uptime (no internet access/customer didn't patch/etc).

I even had a Windows XP laptop in my garage 10+ years ago that did nothing but Chrome and RDP client go over 400 days before I tripped on the power cord.

Yes, systems should be patched and all that. But really, anything not forcing updates should be able to do a year.

I agree that current systems should have no problem running for years between reboots, but which for verious reasons would not be a good idea now.

(conversely my current desktop boots so fast that I now shut it down every day. Saves a bit of power, and a clean start to each day feels good.)

But it's important to understand that back in the 90s running a machine for long periods was unusual. 18 months is certainly notable. Of course there were no security patches back then (and largely no need for them because the always-on Internet wouldn't happen for another few years.)

It's hard to understate just how unreliable tech was in the 20th century. Cars broke down a lot. Computers needed rebooting all the time. The risks taken during Mercury, Gemini and Apollo would be considered completely reckless now.

It has ever been thus. In the first world war more pilots died from mechanical failure than from enemy fire.

First we make it. Then we make it fast. Then we make it reliable.

> Yes, systems should be patched and all that. But really, anything not forcing updates should be able to do a year.

You should try Windows. After some days it becomes confused, some things (especially USB stuff) will not be recognized, programs will not load correctly etc. It is an old bug ( i've seen it also in Win 7) and nobody at Microsoft seems to care about it. Of course, a reboot fixes everything.

At work, I only rebooted my laptop for updates, otherwise I just let it sleep.

On the server, especially when running in a VM, Windows can achieve impressive uptime, I've seen one NT 4.0 server with over 5 years without a reboot.

These days, the question is if you want to keep a server up for that long just because you can. When I worked as a Windows admin, I usually rebooted the servers after installing updates.

I reboot my Win10 machine when there's a power outage long enough to run the UPS down. So probably about once a year.

I haven't had the experience you're describing since everything moved over to NT. I.e., Win95/98 did that for me, NT4, Win2000, XP, etc have never exhibited that behaviour for me.

I think this take, as a generalization, is a couple decades out of date.

05:01:37 up 1442 days, 9:43, 1 user, load average: 4.17, 3.60, 3.38

I really should retire this box, if only because I'm not sure it'd actually reboot.

To be fair, I've seen both a Windows NT 4.0 server, and a Debian server (not sure about the exact version) with several years of uptime. But these days, with everything connected to the Internet, one usually wants to install updates regularly, which often requires a reboot. Also, once the OS goes out of support, it's prudent to upgrade or get a new server altogether.

Worked at Novell from 1992 to 2000 in Novell Consulting. I often worked with customers who had multi-year uptime and took it for granted that upgrades would preserve this. The binary on NetWare was NLM, which stood for NetWare Loadable Module. The name indicates the architecture which meant most upgrades only had to reload an NLM instead of reboot the machine.

I worked on a server that had been up 13 years. This old story talks about one that had been up 16 years (https://m.slashdot.org/story/184077).

I also recall an incident with a school where they needed to reboot the server after years and couldn't find where the server was. They had to use some very sophisticated RTT tools to narrow down where it could be, then knocked some holes in the walls to discover it had been sealed behind a new wall for years.

Good times. But also bad. NetWare was best for file and print serving, but the NLMs that ran so fast ran in ring 0. This means that if you were doing more than file and print serving (like email server or some such) updating NLMs was likely to cause a server reboot because a crash in a new NLM would crash the whole server.

I was a big advocate of Linux back then and would passionate evangelize to my fellow consultants and devs. The most frequent response from people was "well, I sure hope Linux eats into windows NT share" to which is reply "Linux will eat into our share first."

Novel then bought SCO in order to create SuperNOS which was going to meld the best of file/print serving (NetWare) with the best of application serving (unix), but they couldn't pull it off in part because they insisted NetWare should be the base/host operating system.

They offloaded SCO but then eventually aquired a Linux variant and made Linux the base/host operating system with netware's file and print serving as another application on the os. But it was too little too late.

I had started my own business by then and had moved completely away from novel products because of some bad-faith moves some business owners at Novel did in dealings with us. (Long story short, we developed a product they had been telling customers couldn't be done. When we showcased the product at Brainshare, it caused a stir enough for the VP of groupware to ask to speak with us in the subsequent week. During the meeting which was with a much larger audience than I had expected, they kept asking us greater and greater detail how we created the software. The meeting ended abruptly and we had no contact for a few days. I then got a call from a former coworker who was with the product team at the time (he's now a successful web comic artist) and he told me "I don't think what they are doing is right, but I thought I'd let you know. They were going to just ignore you phone calls. Based on what we got out of meeting with y'all, we've decided to built the functionality ourselves and are prepping an announcement to our customers that this functionality will be coming for free soon. That dried up our sales and we decided to pivot completely away from Novel products into Linux products.

Of course all of this is from my memory and I might have some stuff wrong, but this how I remembered things going down. Netware was a great product. Novel made some other great products, but like any other human-led endeavor, it was corruptible to the failings of bias, hubris, and tunnel-vision. C'est la vie.

I was a CNE in the ‘90s and remember fellow CNEs with their uptime screen pictures at the conferences. I was an outlier and usually would do a weekly bounce on my servers. I didn’t necessarily have a good reason to do the bounces, but did not tend to have some of the occasional stray issues others would report. Probably was more superstition than anything.