> But I think it'd be a tough sell to any business to have their users presented with a certificate error message for up to four days because Globalsign screwed up.
Are you making the argument that a business would want the ability to turn off HTTPS enforcement temporarily in case something like the GlobalSign thing happens? Because I think that's a terrible idea. They should just switch to a different CA, or get the CA to sign a certificate that's not impacted by that issue (which I believe GlobalSign offered). Even excluding Let's Encrypt, the price tag for that is < $10 and it's done in a matter of minutes (and only getting better with ACME). Since we're talking about businesses here, better yet: Have a backup certificate from a different CA ready. This is really no different than all the other things you'll need to do to run a reliable service (like have more than one web server). It's not like one CA screwing up means your domain is dead until they've resolved the issue. Same thing goes for HPKP, since a backup pin is required.
Your argument is also not specific to HSTS/HPKP - you're basically saying that HTTPS in general is dangerous for a business because of this. If your site offers HTTPS without HSTS and without redirecting to HTTPS by default, you'd still have users who bookmark HTTPS links, or search engines indexing those links and all of those would also fail, with no way for you to fix it other than switching CAs.
> This is yet another problem that could be solved with DNS signing -- both the risks of HSTS and HPKP would be gone.
I don't want to get into yet another DNSSEC discussion (that's what we have @tptacek for, anyway), but I don't see a huge difference here. Switching CAs can be done in less time than the TTL you're suggesting. Practically speaking, if you want to factor in "time to find a new CA" or something like that, I'd argue that most people won't run their own DNSSEC infrastructure and rather use something like Cloudflare, so the same "time to find a new <foo>" principle would apply here if they've messed up in a way that takes a long time to resolve.
Are you making the argument that a business would want the ability to turn off HTTPS enforcement temporarily in case something like the GlobalSign thing happens? Because I think that's a terrible idea. They should just switch to a different CA, or get the CA to sign a certificate that's not impacted by that issue (which I believe GlobalSign offered). Even excluding Let's Encrypt, the price tag for that is < $10 and it's done in a matter of minutes (and only getting better with ACME). Since we're talking about businesses here, better yet: Have a backup certificate from a different CA ready. This is really no different than all the other things you'll need to do to run a reliable service (like have more than one web server). It's not like one CA screwing up means your domain is dead until they've resolved the issue. Same thing goes for HPKP, since a backup pin is required.
Your argument is also not specific to HSTS/HPKP - you're basically saying that HTTPS in general is dangerous for a business because of this. If your site offers HTTPS without HSTS and without redirecting to HTTPS by default, you'd still have users who bookmark HTTPS links, or search engines indexing those links and all of those would also fail, with no way for you to fix it other than switching CAs.
> This is yet another problem that could be solved with DNS signing -- both the risks of HSTS and HPKP would be gone.
I don't want to get into yet another DNSSEC discussion (that's what we have @tptacek for, anyway), but I don't see a huge difference here. Switching CAs can be done in less time than the TTL you're suggesting. Practically speaking, if you want to factor in "time to find a new CA" or something like that, I'd argue that most people won't run their own DNSSEC infrastructure and rather use something like Cloudflare, so the same "time to find a new <foo>" principle would apply here if they've messed up in a way that takes a long time to resolve.