An insane amount of money and overtime went into changing software and data to make sure that a lot of bad things did not happen. It’s not that the Y2K bug was a nothing burger, a lot of people worked very hard to make sure critical systems were changed.
Yep, now we have the 2038 year coming for Linux. It already got me, I didnt want to renew my home NAS certificate every year, so I thought I’d do a 30 year cert. Well after 2038 it rolled the date to the 1960s…
For your home NAS, I’d recommend using Let’s Encrypt with Certbot. You can use it for internal systems, as long as you have a real domain name. Use DNS verification instead of HTTP. Renewal isn’t an issue if it’s entirely automated.
Many people (me included) like the appeal of a self signed cert in a small homelab. You basically get certificate pinning for free after you trust the cert on all clients.
With your idea, you either have to list a local IP in your public DNS record, or highjack your local DNS to point to the local IP. Both feel inelegant. And you have to give your NAS write access to your API key of your DNS registrar
With your idea, you either have to list a local IP in your public DNS record, or highjack your local DNS to point to the local IP. Both feel inelegant
The DNS records for your internal servers don’t have to be public - they can be only on an internal DNS server if you want to do that. Only the _acme-challenge subdomain has to be public. Let’s Encrypt does follow CNAMEs.
And you have to give your NAS write access to your API key of your DNS registrar
You can use a separate DNS server just for Let’s Encrypt, as it follows CNAMEs. I use acme-dns for this. Let’s Encrypt supports IPv6-only DNS servers so I have my acme-dns instance listening on an IPv6 address in the /64 range on one of my VPSes.
+1 to let’s encrypt and certbot, but pro tip: remember to actually set up certbot, or your friends will laugh at you when your systems all break 6 months later…
So when some Linux apostle is preaching how I need the salvation of Linux in my life, I’ll just tell them that I’m waiting for 2038, and then I’ll jump in AFTER the apocalypse.
Join now lest ye not be saved. 😀 32 bit time keeping is the issue, most systems are 64 bit now, so its just logistics / implementation issue now, not a technology problem
I have been telling people this for 26 years now to no avail. I wish I hadn’t busted my ass now so all the motherfuckers since then who claimed IT is useless could eat the giant dick of downtime.
As it always goes, they only acknowledge you when your actually fixing problems. The work you did that made everything work as it should was never acknowledged the way it should have been.
And it’s tough to remember just how fast computing was changing in the '90s, improving by leaps and bounds all the time with seemingly no ceiling in sight. Consumer computing power was doubling every one and a half years. And in, say, 1994 it wasn’t unreasonable at all to assume that all of that crusty old tech from the '80s and even early '90s surely would have been replaced by the year 2000 anyway without anyone having to do anything special about it. Probably more than once… right?
The crucial disconnect there was that tech people are not necessarily business people and I think a lot of folks grossly underestimated management’s recalcitrance in spending money until it was more than clear they were facing a crisis.
We haven’t changed. Companies will not spend more than they have to on IT if they think they can deal with it until next quarter. This was no different, plus developers of software didn’t expect their stuff to become legacy and not updated with better programs. Memory was premium, so a few less bytes here and there that would work fine for a few years was what they did.
Half a century is a bit of a stretch, but I otherwise agree. It should not have gotten to the level of trouble it became, but I also dislike the implication in the OP that it was just a non-issue meant to scare people; it was a problem that indeed came to a head because many companies kicked the proverbial can, but a potential real problem nonetheless (especially in medical/insurance/monetary systems rather than “planes will rain from the sky” sorts of issues).
An insane amount of money and overtime went into changing software and data to make sure that a lot of bad things did not happen. It’s not that the Y2K bug was a nothing burger, a lot of people worked very hard to make sure critical systems were changed.
Yep, now we have the 2038 year coming for Linux. It already got me, I didnt want to renew my home NAS certificate every year, so I thought I’d do a 30 year cert. Well after 2038 it rolled the date to the 1960s…
Debian is ready - as of Debian Trixie (released in August 2025), all software in the official repo is being compiled with 64-bit time. https://wiki.debian.org/ReleaseGoals/64bit-time
For your home NAS, I’d recommend using Let’s Encrypt with Certbot. You can use it for internal systems, as long as you have a real domain name. Use DNS verification instead of HTTP. Renewal isn’t an issue if it’s entirely automated.
Many people (me included) like the appeal of a self signed cert in a small homelab. You basically get certificate pinning for free after you trust the cert on all clients.
With your idea, you either have to list a local IP in your public DNS record, or highjack your local DNS to point to the local IP. Both feel inelegant. And you have to give your NAS write access to your API key of your DNS registrar
The DNS records for your internal servers don’t have to be public - they can be only on an internal DNS server if you want to do that. Only the
_acme-challengesubdomain has to be public. Let’s Encrypt does follow CNAMEs.You can use a separate DNS server just for Let’s Encrypt, as it follows CNAMEs. I use acme-dns for this. Let’s Encrypt supports IPv6-only DNS servers so I have my acme-dns instance listening on an IPv6 address in the /64 range on one of my VPSes.
+1 to let’s encrypt and certbot, but pro tip: remember to actually set up certbot, or your friends will laugh at you when your systems all break 6 months later…
Sadly the 32bit NAS is stuck at Wheezy, Jessie if you mess around, as the kernel is too big otherwise.
So when some Linux apostle is preaching how I need the salvation of Linux in my life, I’ll just tell them that I’m waiting for 2038, and then I’ll jump in AFTER the apocalypse.
Join now lest ye not be saved. 😀 32 bit time keeping is the issue, most systems are 64 bit now, so its just logistics / implementation issue now, not a technology problem
I have been telling people this for 26 years now to no avail. I wish I hadn’t busted my ass now so all the motherfuckers since then who claimed IT is useless could eat the giant dick of downtime.
As it always goes, they only acknowledge you when your actually fixing problems. The work you did that made everything work as it should was never acknowledged the way it should have been.
And it’s tough to remember just how fast computing was changing in the '90s, improving by leaps and bounds all the time with seemingly no ceiling in sight. Consumer computing power was doubling every one and a half years. And in, say, 1994 it wasn’t unreasonable at all to assume that all of that crusty old tech from the '80s and even early '90s surely would have been replaced by the year 2000 anyway without anyone having to do anything special about it. Probably more than once… right?
The crucial disconnect there was that tech people are not necessarily business people and I think a lot of folks grossly underestimated management’s recalcitrance in spending money until it was more than clear they were facing a crisis.
A lot of stupid businesses and government entities waited until the last fucking second to fix a problem they knew about for half a century.
The overtime should have been exponential for them kicking the can down the road for literal generations.
We haven’t changed. Companies will not spend more than they have to on IT if they think they can deal with it until next quarter. This was no different, plus developers of software didn’t expect their stuff to become legacy and not updated with better programs. Memory was premium, so a few less bytes here and there that would work fine for a few years was what they did.
Half a century is a bit of a stretch, but I otherwise agree. It should not have gotten to the level of trouble it became, but I also dislike the implication in the OP that it was just a non-issue meant to scare people; it was a problem that indeed came to a head because many companies kicked the proverbial can, but a potential real problem nonetheless (especially in medical/insurance/monetary systems rather than “planes will rain from the sky” sorts of issues).