Hacker News new | ask | show | jobs
by jcgl 166 days ago
Looks like it relies on certificate transparency logs. That means that it won’t be monitor endpoints using wildcard certs. Best thing it could do would be to alert when a wildcard cert is expiring without a renewed cert having been issued.
2 comments

Is that enough though? You may have wildcards on domains that are not even on a public DNS and you may forget to replace it "somewhere". For that reason it is better to either dump list of domains from your local DNS or have e.g. zabbix or another agent on every host machine checking that file for you.
That's exactly my point. Is that while this service sounds quite useful for many common cases, it's going to fail in cases where there's not a 1-to-1 certificate-to-server mapping. Even outside of wildcards, you have to account for cases where the cert might be installed on N number of load balancers.
If you're using a cert on multiple IPs, or IPv4+v6, SSLBoard will monitor all IPs. It's not foolproof, but it covers most common practices. btw wildcard certs don't have a good reputation (blast radius)...
I'd say that load balancers (one-address-to-N-servers) count as a common practice, but I otherwise agree in that regard.

Regarding wildcard certs, eh. I wouldn't say they have a bad reputation. Sure, greater blast radius. But sometimes it can certainly simplify things to use one. Your ACME client configuration is easier and your TLS terminator configuration often becomes easier when the terminator would otherwise need to switch based on SNI.

one-address-to-N-servers is perfect if the N servers don't all terminate TLS. If not, it becomes impossible to actually test what certificates are actually served. I've seen this fail before (TLS tests flip/flop between good/bad between checks).

As for wildcard certs, I agree there are use cases where we really need them like dynamic subdomains {customer}.status.com

Can you share how they make ACME client configuration easier?

> Can you share how they make ACME client configuration easier?

It's not a profound difference, but you don't need to add each name to your config. Depending on the team's tooling and processes, that may be inconsequential. But in a setting where config management isn't handled super well, where the TLS terminator is a resource shared by multiple, distinct teams, this is a simplification that can make a difference at the margin.

Think less Cloudflare-scale, and more SMB scale (especially in a Windows shop or recovering Windows shop with a different kind of technical culture than what we might all be implicitly imagining).

Indeed, SSLBoard is scanning CT logs. You can add/import host names though, to allow monitoring of wildcard certs. Same if you're using ports that are not 443, you have to add these to the list of hostnames that are checked.

It's not as convenient, but it's the best SSLBoard can do...