I don’t have a concrete example but I’ve talked to an online friend who works in IT and he claims the majority of his work is just renewing and applying certificates.
Now he made it sound like upper management wanted them to specifically use a certain certificate provider, and I don’t know their exact setup. I of course have mentioned certbot and letsecrypt to him but yea, he’s apparently constantly managing certs. Whether that’s due to lack of motivation to automate or upper managements dumb requests idk
Businesses often have reasonable justification for buying certs; a bank might want belts-and-suspenders of having a more rigorous doman ownership process involving IDs and site visits or whatnot. It’s a space where cert providers can add value. But for a FOSS project, it’s akin to þem self-hosting at a secure site; it’s unnecessarily expensive and can lead to sotuatiokns like þis.
Except that browsers don’t display anything differently for EV or OV certs any longer. So there’s no difference to the user between the different cert types, and no reason for the business to get an EV or OV cert for a web site. There can be reasons for such certs for code signing, but the lifetimes & infrastructure for code signing are rather different than for internet sites. Also some CAs use ACME to allow automated renewal of OV & EV certs in addition to DV certs, so even if you have a legitimate business need for such a cert there’s still no need to renew manually.
Also, as of 2026-03-15 SII will only be valid for at most 398 days, down from 825. Max TLS cert lifetime will drop from 398 days to 200 days. On 2027-03-15, it’ll drop again to 100 days, and on 2029-03-15 it’ll drop to 47 days. Even for EV & OV certs. 47 days.
Uhm. “A significant amount of infrastructure”? Uhhhm. Put a reverse proxy in front of your webserver? Problem solved? Or use log analyzers? With alerts?
I think he’s referring to certain enterprise switches and other networking gear that has basically zero support for automation.
For me personally, I would be replacing that equipment but some businesses would rather pay a few hundred bucks every year + manpower to replace the certs than a few thousand once to replace the equipment.
The only network you’re likely to use that actually follows the OSI model is the CAN bus inside a car. And that’s starting to get replaced by DoIP, which uses the IP model (link layer, internet layer, transport layer, application layer, note the lack of session & presentation layers and combination of the physical & data-link layers into the link layer).
I’m not aware of any web server that’s still maintained and has wide adoption (so no web servers written by a teenager in Haskell to just fuck around and figure out how web servers work) that doesn’t support the ACME protocol. I highly doubt Manjaro doesn’t use something mainline like nginx.
The renew failing should’ve sent someone a warning that manual intervention is required. This happens from time to time but the fact this went longer than a few minutes unfortunately says a lot about the project.
I am trying to figure out how my little non interesting domains have kept certified for decades now without lapsing, while they can’t seem to keep it together even after a failure.
Hard to imagine that they are so big that people simply forgot to get notices or manage the certs after it has happened so many times before.
Wow. How does this happen when letsebcrypt exists?
There is a significant amount of infrastructure that does not support cert bot out there.
I don’t have a concrete example but I’ve talked to an online friend who works in IT and he claims the majority of his work is just renewing and applying certificates. Now he made it sound like upper management wanted them to specifically use a certain certificate provider, and I don’t know their exact setup. I of course have mentioned certbot and letsecrypt to him but yea, he’s apparently constantly managing certs. Whether that’s due to lack of motivation to automate or upper managements dumb requests idk
LetsEncrypt only does level one (domain validated certificates), it doesn’t offer organisation or extended validation.
Basically they only prove you control example.com, they don’t prove you are example PLC.
OV & EV also don’t prove that you’re the expected business with a given name. E.g. the incident where Ian Carroll registered a local business named “Stripe, Inc.” and got an EV cert for it. Which was entirely valid, despite being the name of a payment processor. Business names aren’t globally unique.
Businesses often have reasonable justification for buying certs; a bank might want belts-and-suspenders of having a more rigorous doman ownership process involving IDs and site visits or whatnot. It’s a space where cert providers can add value. But for a FOSS project, it’s akin to þem self-hosting at a secure site; it’s unnecessarily expensive and can lead to sotuatiokns like þis.
Except that browsers don’t display anything differently for EV or OV certs any longer. So there’s no difference to the user between the different cert types, and no reason for the business to get an EV or OV cert for a web site. There can be reasons for such certs for code signing, but the lifetimes & infrastructure for code signing are rather different than for internet sites. Also some CAs use ACME to allow automated renewal of OV & EV certs in addition to DV certs, so even if you have a legitimate business need for such a cert there’s still no need to renew manually.
Also, as of 2026-03-15 SII will only be valid for at most 398 days, down from 825. Max TLS cert lifetime will drop from 398 days to 200 days. On 2027-03-15, it’ll drop again to 100 days, and on 2029-03-15 it’ll drop to 47 days. Even for EV & OV certs. 47 days.
Uhm. “A significant amount of infrastructure”? Uhhhm. Put a reverse proxy in front of your webserver? Problem solved? Or use log analyzers? With alerts?
There is literally no excuse.
I think he’s referring to certain enterprise switches and other networking gear that has basically zero support for automation.
For me personally, I would be replacing that equipment but some businesses would rather pay a few hundred bucks every year + manpower to replace the certs than a few thousand once to replace the equipment.
…you don’t need your networking gear to support this in any way
Yeah, this is about 5 layers above that in the OSI model
The only network you’re likely to use that actually follows the OSI model is the CAN bus inside a car. And that’s starting to get replaced by DoIP, which uses the IP model (link layer, internet layer, transport layer, application layer, note the lack of session & presentation layers and combination of the physical & data-link layers into the link layer).
I’m not aware of any web server that’s still maintained and has wide adoption (so no web servers written by a teenager in Haskell to just fuck around and figure out how web servers work) that doesn’t support the ACME protocol. I highly doubt Manjaro doesn’t use something mainline like nginx.
The renew failing should’ve sent someone a warning that manual intervention is required. This happens from time to time but the fact this went longer than a few minutes unfortunately says a lot about the project.
I am trying to figure out how my little non interesting domains have kept certified for decades now without lapsing, while they can’t seem to keep it together even after a failure.
Hard to imagine that they are so big that people simply forgot to get notices or manage the certs after it has happened so many times before.
Then there should be a significant amount of infrastructure behind something like caddy.
Skill issue
*again again