I wish someone could do a genuinely free SSL CA with "reasonable" procedures, just to disrupt the industry. Charging $19 to 500/yr for something which is fundamentally non-rival (unlike, say, a domain name) with very minimal marginal cost is basically bullshit rent-seeking.
SSL certs are an utter commodity, and generally the CA system is a broken system, but we'd be a lot better off with free "real" SSL certs everywhere rather than a bunch of self signed stuff. The loss here is the large number of applications with self-signed certs on the Internet, as well as the general hassle involved in purchasing and renewing certs for everyone who does.
The cost of having a well-run base CA is small number of millions to set up and maybe $1-2mm/yr. It would do a lot more good than many other charities. The marginal costs to do mail-from auth S/MIME and domain-verify certs would be limited.
I used to think StartCom was a good solution for this, but after recent staff departures, and their shortsighted-at-best policy here, I can't recommend them.
With commercial CAs, people use wildcard certs in a lot of places where they should be using distinct other certs. It's maybe not an issue for https for main company websites ($49/yr isn't too big a deal), but I want people to use real certs for START TLS email, client certs, etc. I also want one CA to be really popular, or at least one simple cert-issuance protocol to be popular, so "automatically generate a cert and CSR and get the cert loaded" becomes an automatic part of software setup.
I'm kind of surprised Google hasn't done this, in exchange for people proving ownership of a Google+ account or something. FB, Twitter, etc. could all easily bury the costs of a genuinely free CA.
I want people to use real certs for START TLS email, client certs, etc.
Because StartSSL issues them for free, I actually did exactly that -- used a separate certificate for each of my secure services (IMAP, SMTP, HTTPS, etc.), on each of my domains. I believe I currently have roughly 25-30 certificates issued by them. My thinking was that, despite the extra configuration complexity, if a particular key were compromised, at least I wouldn't have to replace all my certificates!
Ugh.
So now, ironically, I'm faced with having to replace all my certificates, not to mention paying StartSSL $25 per certificate if I want to do the right thing again and have all the existing ones revoked, as well.
Instead, I'm tempted to "pay up" for a wildcard certificate for each of my domains from a provider whose revenue stream comes from minting certs, not revoking them. Configuring my services sure would be easier with 5 certificates rather than 25. Also, I've never leaked a key by any error of my own, and I'm beginning to think that's less likely than someone finding another hole in OpenSSL (or a CA) which requires wholesale certificate replacement again.
Maybe, but this would add legitimacy to X.509 and the CA business and cement it even further. It's an enormous ongoing drag to Internet security.
I'd like to see Mozilla and Google to team up on a reasonable replacement. Even Microsoft might join up given its recent shifts in attitude. TLS supports non-X.509 stuff just fine.
'per year' pricing is one of the most insidious aspects of the current CA system.
Certificates don't rot.
They claim that annual renewal is necessay to protect us from a rogue but unrevoked certificate; however any malicious activities would be quite profitable well within a year.
So why not set renewal to be monthly, just to be extra-safe? Weekly? Perhaps I should suggest that to them. The resulting outcry from users might be the only way we can disrupt the CA situation.
cacert.org is trying just that for a long time. The hard part is the paperwork and procedures that give OS and browser vendors the trust to add you in their default config.
Well, that is with good reason. A rogue or insecure root CA which is trusted by major browsers/OS'ses would basically invalidate the whole CA system (see the DigiNotar affair from 2011).
Comments
I wish someone could do a genuinely free SSL CA with "reasonable" procedures, just to disrupt the industry. Charging $19 to 500/yr for something which is fundamentally non-rival (unlike, say, a domain name) with very minimal marginal cost is basically bullshit rent-seeking.
SSL certs are an utter commodity, and generally the CA system is a broken system, but we'd be a lot better off with free "real" SSL certs everywhere rather than a bunch of self signed stuff. The loss here is the large number of applications with self-signed certs on the Internet, as well as the general hassle involved in purchasing and renewing certs for everyone who does.
The cost of having a well-run base CA is small number of millions to set up and maybe $1-2mm/yr. It would do a lot more good than many other charities. The marginal costs to do mail-from auth S/MIME and domain-verify certs would be limited.
I used to think StartCom was a good solution for this, but after recent staff departures, and their shortsighted-at-best policy here, I can't recommend them.
With commercial CAs, people use wildcard certs in a lot of places where they should be using distinct other certs. It's maybe not an issue for https for main company websites ($49/yr isn't too big a deal), but I want people to use real certs for START TLS email, client certs, etc. I also want one CA to be really popular, or at least one simple cert-issuance protocol to be popular, so "automatically generate a cert and CSR and get the cert loaded" becomes an automatic part of software setup.
I'm kind of surprised Google hasn't done this, in exchange for people proving ownership of a Google+ account or something. FB, Twitter, etc. could all easily bury the costs of a genuinely free CA.
I want people to use real certs for START TLS email, client certs, etc.
Because StartSSL issues them for free, I actually did exactly that -- used a separate certificate for each of my secure services (IMAP, SMTP, HTTPS, etc.), on each of my domains. I believe I currently have roughly 25-30 certificates issued by them. My thinking was that, despite the extra configuration complexity, if a particular key were compromised, at least I wouldn't have to replace all my certificates!
Ugh.
So now, ironically, I'm faced with having to replace all my certificates, not to mention paying StartSSL $25 per certificate if I want to do the right thing again and have all the existing ones revoked, as well.
Instead, I'm tempted to "pay up" for a wildcard certificate for each of my domains from a provider whose revenue stream comes from minting certs, not revoking them. Configuring my services sure would be easier with 5 certificates rather than 25. Also, I've never leaked a key by any error of my own, and I'm beginning to think that's less likely than someone finding another hole in OpenSSL (or a CA) which requires wholesale certificate replacement again.
You can't use wildcard for EV domains.
Thankfully, I don't need those!
Mozilla comes to mind as an organization that could be trusted to do this.
Maybe, but this would add legitimacy to X.509 and the CA business and cement it even further. It's an enormous ongoing drag to Internet security.
I'd like to see Mozilla and Google to team up on a reasonable replacement. Even Microsoft might join up given its recent shifts in attitude. TLS supports non-X.509 stuff just fine.
Why does it have to be free? 5 bucks a year for an SSL cert is peanuts.
If you want free, push for killing CAs :)
Charging any amount of money adds friction. $0 to $5 is a bigger jump than $5 to $50. All I care about is having SSL everywhere.
Maybe someone like CloudFlare could cover the "Free CA" project.
'per year' pricing is one of the most insidious aspects of the current CA system.
Certificates don't rot.
They claim that annual renewal is necessay to protect us from a rogue but unrevoked certificate; however any malicious activities would be quite profitable well within a year.
So why not set renewal to be monthly, just to be extra-safe? Weekly? Perhaps I should suggest that to them. The resulting outcry from users might be the only way we can disrupt the CA situation.
I think the billing interval is relatively unimportant.
cacert.org is trying just that for a long time. The hard part is the paperwork and procedures that give OS and browser vendors the trust to add you in their default config.
Well, that is with good reason. A rogue or insecure root CA which is trusted by major browsers/OS'ses would basically invalidate the whole CA system (see the DigiNotar affair from 2011).
The CNNIC root is still trusted by most browsers/OSes.
Sure, but it goes counter to "I wish someone could do a genuinely free SSL CA with "reasonable" procedures".
So bad news: it's hard. good news: people are trying.
More bad news: They're failing pretty hard. CACert have been around for a long time, and still haven't managed to get themselves audited.
There was an interesting thread on the subject on the crypto-list last year ("How much does it cost to start a root CA ?"), see eg:
http://lists.randombit.net/pipermail/cryptography/2013-Janua...
http://lists.randombit.net/pipermail/cryptography/2013-Janua...
http://lists.randombit.net/pipermail/cryptography/2013-Janua...
And for good measure, on the subject of certs and trust, the thread after:
"another cert failure" (2011)
http://lists.randombit.net/pipermail/cryptography/2013-Janua...