If you like this sort of thing, perhaps you'll enjoy my SSL/TLS and PKI history where I track a variety of ecosystem events starting with the creation of SSL in 1994: https://www.feistyduck.com/ssl-tls-and-pki-history/
Does the TLS working group know that? Pretty much all their design work apart from one guy representing email use via Postfix and a few guys working on a perpetual-motion-machine profile for telcos assumes the only possible use for TLS is the web.
The short lived certificates started making a lot more sense to me when I discovered I could get Let's Encrypt to issue IP address certs. Clearly, in this context of use we need our certificates to die quickly.
You can now make any web server operate with a publicly valid TLS certificate without paying any money, registering a domain, configuring DNS or disclosing any personally identifiable information. It can be entirely automatic and zero configuration. The only additional service required is something like a STUN server so the public IP can be discovered and updated over time.
I am reading your comment and find the proposition interesting, but I can't quite understand the part about the STUN server - doesn't that "just" help me find my own public IP address ? Do you mean that I could then give out this address to others (instead of them having to do a DNS lookup) so they can connect to the webserver ?
> I am reading your comment and find the proposition interesting, but I can't quite understand the part about the STUN server - doesn't that "just" help me find my own public IP address ?
He is hosting his domain on a machine behind a reverse proxy over which he has no control (common enough); in this case the server will not know its own public IP as all resolves to (for example) `www.mydomain.com` will return the address of the proxy. To get the public IP he uses a STUN (or similar) public-facing service.
Not quite sure why he needs the public IP, though: from what I remember, the certs include the domain, not the IP.
You can issue a TLS certificate with a SAN that is a literal IPv4 address. You do not need a domain to serve TLS to clients. It definitely helps with the UX, but it's not mandatory for the browsers and other web tech to function.
If you're running private PKI, sure, you'll do it.
What value is it when you are behind a proxy that can change IP? I mean, I'm going on the assumption that the proxy is not under his control, nor does it do the tls termination.
If your public interface address can change, it does dramatically reduce the value of a purely IP-addressed host. But I don't think it eliminates it entirely.
With a dynamic IP you can still detect a change, reissue a cert for the new IP and proceed automatically. There are self-hosting and machine-to-machine scenarios where this amount of autonomy could be welcome.
Yes the point is to simply discover the public ip you present as on the internet. It's not a particularly hard problem to solve, but you often can't know your public interface just from inside the machine. Being behind a NAT with TCP 80/443 forwarded to the actual web server is an example.
It's worth noting that while splitting the PKI hierarchies is a good thing, the CABF does provide rules S/MIME (email signing) and Code Signing. Also, "WebPKI" never actually appears in the BR documents from what I can see, nor do they require the use of HTTP (hence why you can use these for SMTP).
There's a huge suggestion in here which would make PKI vastly more respectable: Disallowing root programs (browser operators) from also being CAs. I loudly suggested at the time Google Trust Services should be rejected but the Mozilla rep loudly defended approving a CA from a root program from a company that happens to pay their entire salary.
PKI as it stands is only a few steps from Google just deciding everyone must have a short-lived certificate from Google to be on the web.
While I sort-of see what you're trying to say, if you knew the groups and teams involved - you'd know there was no favouritism and a strong degree of separation between CA and root programs.
The root programs who have their own CAs are also cloud providers, who arguably have a legitimate need for the CA. Or in Apple's case they have their own CA, but don't issue externally. They keep CA and root program separate.
That's absolutely incorrect. While CABF sets the 'Baseline Requirements' that ultimately go into the WebTrust audit scheme that root programs use to accept roots into their trust stores...browsers can and do set their own rules.
The reduction of TLS cert lifetime to a max of 398 days was an Apple policy.
Here's a link to the minutes of the CABF meeting where the 25 certificate issuers and the 4 browser vendors—Apple, Google, Microsoft, Mozilla—agreed to reduce the validity period of TLS certificates unanimously [1].
> The reduction of TLS cert lifetime to a max of 398 days was an Apple policy.
Actually, all of the browser vendors voted to reduce the validity period of TLS certificates from 825 days to 398 days at the September 2019 meeting. The ballot failed because a majority of the certificate issuers voted against it.
At the February 2020 CABF meeting, Apple announced it would unilaterally enforce the 398-day limit through its own root program policy. Starting September 1, 2020, any new TLS certificate with a validity period exceeding 398 days would simply not be trusted by Safari, macOS, or iOS.
This effectively made the 398-day limit a de facto standard — no CA would issue longer certificates if they’d be rejected by Apple devices [2].
|Date |Max Certificate Validity|SAN Data Reuse Period|
|-----------------|------------------------|---------------------|
|Before March 2026|398 days (current) |398 days |
|March 15, 2026 |200 days |200 days |
|March 15, 2027 |100 days |100 days |
|March 15, 2028 |47 days |10 days |
|March 15, 2029 |47 days |10 days (final) |
That's fundamentally untrue. Browser root programs unilaterally decide their requirements, and also unilaterally boot other CA/B members out of the CA/B. The CA/B is not by any means a useful functional body, it's an excuse why companies can't be forced to undo things after the fact.
You can now make any web server operate with a publicly valid TLS certificate without paying any money, registering a domain, configuring DNS or disclosing any personally identifiable information. It can be entirely automatic and zero configuration. The only additional service required is something like a STUN server so the public IP can be discovered and updated over time.
He is hosting his domain on a machine behind a reverse proxy over which he has no control (common enough); in this case the server will not know its own public IP as all resolves to (for example) `www.mydomain.com` will return the address of the proxy. To get the public IP he uses a STUN (or similar) public-facing service.
Not quite sure why he needs the public IP, though: from what I remember, the certs include the domain, not the IP.
What value is it when you are behind a proxy that can change IP? I mean, I'm going on the assumption that the proxy is not under his control, nor does it do the tls termination.
With a dynamic IP you can still detect a change, reissue a cert for the new IP and proceed automatically. There are self-hosting and machine-to-machine scenarios where this amount of autonomy could be welcome.
PKI as it stands is only a few steps from Google just deciding everyone must have a short-lived certificate from Google to be on the web.
The root programs who have their own CAs are also cloud providers, who arguably have a legitimate need for the CA. Or in Apple's case they have their own CA, but don't issue externally. They keep CA and root program separate.
If you don’t like a particular CA’s policies, you can choose a different one.
The reduction of TLS cert lifetime to a max of 398 days was an Apple policy.
Here's a link to the minutes of the CABF meeting where the 25 certificate issuers and the 4 browser vendors—Apple, Google, Microsoft, Mozilla—agreed to reduce the validity period of TLS certificates unanimously [1].
> The reduction of TLS cert lifetime to a max of 398 days was an Apple policy.
Actually, all of the browser vendors voted to reduce the validity period of TLS certificates from 825 days to 398 days at the September 2019 meeting. The ballot failed because a majority of the certificate issuers voted against it.
At the February 2020 CABF meeting, Apple announced it would unilaterally enforce the 398-day limit through its own root program policy. Starting September 1, 2020, any new TLS certificate with a validity period exceeding 398 days would simply not be trusted by Safari, macOS, or iOS.
This effectively made the 398-day limit a de facto standard — no CA would issue longer certificates if they’d be rejected by Apple devices [2].
[1]: https://cabforum.org/2025/04/11/ballot-sc081v3-introduce-sch...[2]: https://www.entrust.com/blog/2020/02/apple-announces-398-day...