> StartSSL won't give me a cert (they did previously, but want $$$ now)
Do you have reference for that? I've been using their free certs for a few things for a while and a recent renewal was free. I'll be paying for one shortly, but that is only because I want a wildcard for one of my domains in order to save some faf.
I suspect your problem is the "third level domain" thing: they would effectively need to have permission from the over of the next level up to sign a cert on your behalf for one of their names (or you would need your domain provider to take part in the validation process).
> for those that can afford it - the rest of us doomed to plaintext or rejected as "self signed"
That certainly isn't the intention of the security community, but until a method is found that allows certificates to be freely produced without the security risks involved with self-signed certs that is what we (well, those in the poverty trap) are stuck with.
How StartSSL arrange their business model seems to be one solution, your current problem above not withstanding. If the depth of your free domain is the issue then perhaps the solution lies with the domain registrant taking the cost of a wildcard certificate (~$30/yr with startssl, ~$60/yr elsewhere) and so their subscribers can use it for SSL. They would have to proxy web requests to you though (with something like nginx in a reverse proxy configuration), so it would not be a zero admin option and would cost them in extra bandwidth requirements too unless they already host your content (rather than just responding to DNS requests for content hosted elsewhere), otherwise the only way you could use the cert is for you to know the private key which breaks the assurance model completely. Perhaps you could convince a provider of third level domains that taking these steps could give them with a competitive advantage over similar providers who do not.
Is there something about a chain of trust people don't understand? A chain of trust doesn't mean a commercial body, as that implies that being commercial makes it more trustworthy.
It means any individuals or groups you can trust, not those your browser maker or software developer includes in their software in the belief that they are trustworthy. Why should relationships with faceless commercial entities whose employees you have no personal acquaintance with take precedence over people known to you believe you can trust?
Why shouldn't individuals be capable of forming the same trust networks that companies do, and what methods do companies use in forming their chains of trust that individuals can't emulate, as companies themselves depend on individual human beings to create their systems?
Did you overlook the last paragraph in the article?
The way the browser vendors, Mozilla being one of the bad guys, notify users about untrusted certificates is a major culprit. They state that the certificate is not trusted, or something to that effect, when in fact they mean that the certificate is not recognized in their trust network.
Correct, but it does mean entities that are qualified to be trusted on the matter. I don't trust a lot of my contacts (both personal or commercial) to be good people for judging what is safe/legit/secure.
How many times I've heard an otherwise technically competent individual complain at being laughed at for using "but <other_person> give me the link I assumed it was OK" as part of their defence when taken the mick out of for getting infected by malware from some ropey website/app/whatever is something it would scare me to count.
> Why shouldn't individuals be capable of forming the same trust networks that companies do
You can. The current system does allow this, though I'll admit not at all intuitively for the general public. Create your own CA keys, have contacts include them in their trust stores. Build a web of trust by signing each others keys (or passing them between each other and new contacts via secure channels).
> They state that the certificate is not trusted, or something to that effect, when in fact they mean that the certificate is not recognized in their trust network.
From their PoV those two thing are the same. Things I have no reason to trust are not things I can guarantee are trustworthy. In security there is unfortunately no room for granting the "benefit of the doubt".
> I suspect your problem is the "third level domain"
That's one of the problems for me at the moment. A friend in a similar situation (but with regular .net domain) triggered one of their "must pay now" criteria (ref: the StartSSL business model).
They have worked in the past, and are sometimes an option, but a "second source" is unfortunately necessary.
> That certainly isn't the intention of the security community
I don't really intend to make any accusations - take any harsh tone in my post to be a general annoyance at the current situation.
I appreciate the ideas about technical workarounds. They are similar to some things I've considered already, and might yet try.
I'm more concerned about a general solution to this PKI/cert problem - XMPP and HTTP simply being the common examples. In reference to the manifesto linked at the top of this thread, the time to encrypt everything is "now". This is a great idea, but it can't happen everywhere until the money-issue needs to be solved.
Your tone was actually fine, I put that there as a preface to the next statement (which was essentially "but that is the way things sometimes work out").
> I appreciate the ideas about technical workarounds.
Another suggestion you might try if you know enough reliable people in a similar position as you find yourself in (happy to have a third level domain but needing a certificate potentially for commercial use) who you trust enough to do business with: club together, and rent a normal domain and wildcard certificate with it.
Of course someone will have to take responsability handling the money, for technical admin, and for enforcing certain rules such as "nothing illegal otherwise we are all at risk" (this person will need to be trusted by the other parties, it goes without saying), and if a two-or-three $ per year is the maximum you can stretch to each then you would need a small number of tens of people in (more if you can each spare less) (domain + wildcard from StartSSL ~= $75/2yr ~= $2/year each for 20 people, still not free but an order of magnitude closer), and there is the support "risk" of not particulary technically minded peopel joining and needing hekp to setup/maintain things, so the idea might be completely impractical but nonetheless there is a chance it might be worth considering.
> A friend in a similar situation (but with regular .net domain) triggered one of their "must pay now" criteria (ref: the StartSSL business model).
One thing that surprises some who have been using SlartSSL for a some time is that a while ago a "non commercial use only" clause was added to the free certificates. I don't know how they police that if they do at all, but if your friend's site was taking money for services that may explain the change from his/her account's PoV. I've stopped using the free certificates for a couple of sites I administer for that reason. This doesn't apply to any of the other certificates they offer (though for a single domain it is cheaper to buy elsewhere).
I've not used CACert since investigating a few years ago. At that time it was the case that their CA certs were not trusted by many platforms (i.e. most Windows users) so if you have users on those platforms their certificates are no better than self-signed ones for those users - this made them unacceptable for what I was doing at the time. If http://en.wikipedia.org/wiki/CAcert.org is up-to-date then this has not changed.
StartSSL's certificates are accepted by all major platforms (including all Windows desktop variants from XP upwards since ~2009), though there are gaps in this coverage. The stock browser and mail client in at least some relatively recent variants of Windows Phone didn't like them last time I tried. They should be trusted by Android since 2.2, though I believe there are issues with some builds of 2.2/2.3.x (I've not heard of trouble with later versions).
StartSSL's interface can be quite clunky but I've never had it fail. Certificates usually come through in good time (often faster than certain paid certificates I could mention). The "non commercial use" clause added to the terms for the free certificates in recent times may be a problem for some. As I've not really used CACert I can't comment on how they compare on these matters.
They are, but they're not included in various browsers for various reasons -- so they are "legit", but they are not easy to use for use-cases where you don't have a modicum of control over clients (can install, or ask clients to install, cacert root keys).
Please don't suggest that cacert is much less secure than trusting a handful of government CAs by default (or even much less secure than certain commercial CAs).
Cacert isn't perfect, but it is an interesting and important project. It's a pity Debian ended up stripping cacert IMNHO. Anyway, it is healthy to be sceptical, for some more info, see eg:
Do you have reference for that? I've been using their free certs for a few things for a while and a recent renewal was free. I'll be paying for one shortly, but that is only because I want a wildcard for one of my domains in order to save some faf.
I suspect your problem is the "third level domain" thing: they would effectively need to have permission from the over of the next level up to sign a cert on your behalf for one of their names (or you would need your domain provider to take part in the validation process).
> for those that can afford it - the rest of us doomed to plaintext or rejected as "self signed"
That certainly isn't the intention of the security community, but until a method is found that allows certificates to be freely produced without the security risks involved with self-signed certs that is what we (well, those in the poverty trap) are stuck with.
How StartSSL arrange their business model seems to be one solution, your current problem above not withstanding. If the depth of your free domain is the issue then perhaps the solution lies with the domain registrant taking the cost of a wildcard certificate (~$30/yr with startssl, ~$60/yr elsewhere) and so their subscribers can use it for SSL. They would have to proxy web requests to you though (with something like nginx in a reverse proxy configuration), so it would not be a zero admin option and would cost them in extra bandwidth requirements too unless they already host your content (rather than just responding to DNS requests for content hosted elsewhere), otherwise the only way you could use the cert is for you to know the private key which breaks the assurance model completely. Perhaps you could convince a provider of third level domains that taking these steps could give them with a competitive advantage over similar providers who do not.