5 ms·

Distrust Let's Encrypt Subordinate Certificate

37 points by _jomo 11y ago

tptacek 11y ago
This is just --- apologies for being coarse about it --- some rando filing a bug, right? I looked and didn't see any evidence that he's somehow associated with the Firefox project.
- onli 11y ago
  If I see that right, it could be the Geschäftsführer (CEO) from https://www.psw-group.de/ https://www.psw-group.de/. I saw no further Mozilla involvement in my short search either.
  - tptacek 11y ago
    Which is a CA reseller. Nice.
    - userbinator 11y ago
      No doubt a lot of other CAs are going to be pretty perturbed once LE really gets widespread. What reason for existence will they have then? EV certs?
      - ceejayoz 11y ago
        GoDaddy claims support is a selling point. Which, if you've used GoDaddy's support, is hilariously laughable.
      - dingaling 11y ago
        > What reason for existence will they have then? EV certs? I doubt that corporate users will move in any significant numbers to LE, even for basic DV certs. Two reasons: 1. The certificate renewal process appears to be incompatible with most corporate change-control processes I've encountered: nothing may change on a production server unless an admin is given authorisation to change it. A cron job that could take all customer-facing services offline is just crazy[0]. 2. When something goes wrong with a cert issuance or renewal, corporate users ( want | need ) to be able to call a human to have them fix it, regardless of cost. And even that cost is insignificant in context to revenue. I'm sure LE will be popular with bloggers and hobby users. Beyond that? [0] yes, the admins could run the LE client manually in non-prod and move the certs over to production. But that's no easier than current cert renewals, and would have to occur every 60-90 days.
        toomuchtodo 11y ago
        Someone who is responsible for a site's uptime is going to be more conservative about changes being made, most definitely. As a sysadmin, I don't think Let's Encrypt's 90 day expiration period is too onerous, but if I can sidestep that aggressive expiration period with a $300-600 wildcard cert that expires yearly from a traditional CA, that's the route I'm going to go (when my fully loaded costs to my employer are on the low end of that cost per hour). Disclaimer: I'm a big proponent of Let's Encrypt, and tweeted at Jeff Barr @ AWS asking if they could integrate its lifecycle into AWS' IAM SSL ecosystem.
        userbinator 11y ago
        I think servers are patched a bit more often than 90 days now, so the expiration period is not that short.
        toomuchtodo 11y ago
        Patching servers and cycling SSL certificates are apples and oranges. Depending on the size of the org, you may have several divisions that need to have new key/certs distributed to them, and their environments updated through change requests processes. Not everyone are teams of 5-20 on EC2 or DigitalOcean.
  - jepler 11y ago
    He has posted on at least one other occasion on mozilla's bugzilla, see this 2010 comment on another certificate-related bug: https://bugzilla.mozilla.org/show_bug.cgi?id=556468#c33 https://bugzilla.mozilla.org/show_bug.cgi?id=556468#c33
    - jepler 11y ago
      And also on this bugzilla item, again questioning whether Mozilla is properly following procedure WRT letsencrypt https://bugzilla.mozilla.org/show_bug.cgi?id=1204656 https://bugzilla.mozilla.org/show_bug.cgi?id=1204656
      - onli 11y ago
        This one I saw. I did not count it as mozilla involvement. Nonetheless good to have it linked here to see the train of thought.
jpdus 11y ago
From the comments: A PITRA (point-in-time readiness assessment) is sufficient to begin issuing if the formal audit follows in a timely fashion. It's not possible to have a formal audit of the issuance process unless you are issuing; requiring it to begin issuance would lead to a chicken-and-egg problem. Gerv
- rgbrenner 11y ago
  not really true. Let's encrypt is cross signed with IdenTrust. So whether the ISRG certificate is trusted or not, they can still issue valid certificates.
idibidiart 11y ago
Iran gamed the CA system (DigiNotar?) to spy on their citizen's https traffic. http://www.theregister.co.uk/2011/09/06/diginotar_audit_damning_fail/ http://www.theregister.co.uk/2011/09/06/diginotar_audit_damn... So it's OK to be paranoid, IMO. The CA system is radically flawed in that it can be gamed by state actors.
- rockdoe 11y ago
  Completely and utterly irrelevant to the issue at hand.
  - idibidiart 11y ago
    The issue at hand matters far less than the issue that the issue at hand is part of :) which is the CA system itself. You can try and move the corporatist agenda for a two tier security system where the mighty and powerful have all seeing eyes and the consumer is told that all their https communication is secure and private.
idibidiart 11y ago
The issue at hand matters far less than the issue that the issue at hand is part of, which is the CA system itself. With the CA system, you have a two tier threat model where the mighty and powerful can have all seeing eyes and the consumer is told that all their https communication is secure and private. Certificate Transparency and other proposals may help fix this, but the first thing to do is to acknowledge that there is a problem, and so far the major players have yet to acknowledge the fundamental problems of the current CA system. LetsEncrypt makes it worse by promoting the wider adoption of this flawed system. Certainly won't help keep us secure against the Chinese or anyone with big resource. Update: Stating that there is no solution without any knowledge of the existing solutions is itself part of the problem. Let's start with this: https://github.com/okTurtles/dnschain https://github.com/okTurtles/dnschain
- userbinator 11y ago
  What alternatives are there? SSH-style "you have not visited this site before, do you want to trust it?"? Distributed trust is a hard problem.
  - idibidiart 11y ago
    Stating that there is no solution without any knowledge of the existing solutions is itself part of the problem. Let's start with this: https://github.com/okTurtles/dnschain https://github.com/okTurtles/dnschain
idibidiart 11y ago
The issue at hand matters far less than the issue that the issue at hand is part of, which is the CA system itself. With the CA system, you have a two tier threat model where the mighty and powerful can have all seeing eyes and the consumer is told that all their https communication is secure and private. Certificate Transparency and other proposals may help fix this, but the first thing to do is to acknowledge that there is a problem, and so far the major players have yet to acknowledge the fundamental problems of the current CA system. LetsEncrypt makes it worse by promoting the wider adoption of this flawed system. Certainly won't help keep us secure against the Chinese or anyone with big resource. Update: Stating that there is no solution without any knowledge of the existing solutions is itself part of the problem. Let's start with this: https://github.com/okTurtles/dnschain https://github.com/okTurtles/dnschain

37 points by _jomo 11y ago

This is just --- apologies for being coarse about it --- some rando filing a bug, right? I looked and didn't see any evidence that he's somehow associated with the Firefox project.

onli 11y ago
If I see that right, it could be the Geschäftsführer (CEO) from https://www.psw-group.de/ https://www.psw-group.de/. I saw no further Mozilla involvement in my short search either.
- tptacek 11y ago
  Which is a CA reseller. Nice.
  - userbinator 11y ago
    No doubt a lot of other CAs are going to be pretty perturbed once LE really gets widespread. What reason for existence will they have then? EV certs?
    - ceejayoz 11y ago
      GoDaddy claims support is a selling point. Which, if you've used GoDaddy's support, is hilariously laughable.
    - dingaling 11y ago
      > What reason for existence will they have then? EV certs? I doubt that corporate users will move in any significant numbers to LE, even for basic DV certs. Two reasons: 1. The certificate renewal process appears to be incompatible with most corporate change-control processes I've encountered: nothing may change on a production server unless an admin is given authorisation to change it. A cron job that could take all customer-facing services offline is just crazy[0]. 2. When something goes wrong with a cert issuance or renewal, corporate users ( want | need ) to be able to call a human to have them fix it, regardless of cost. And even that cost is insignificant in context to revenue. I'm sure LE will be popular with bloggers and hobby users. Beyond that? [0] yes, the admins could run the LE client manually in non-prod and move the certs over to production. But that's no easier than current cert renewals, and would have to occur every 60-90 days.
      - toomuchtodo 11y ago
        Someone who is responsible for a site's uptime is going to be more conservative about changes being made, most definitely. As a sysadmin, I don't think Let's Encrypt's 90 day expiration period is too onerous, but if I can sidestep that aggressive expiration period with a $300-600 wildcard cert that expires yearly from a traditional CA, that's the route I'm going to go (when my fully loaded costs to my employer are on the low end of that cost per hour). Disclaimer: I'm a big proponent of Let's Encrypt, and tweeted at Jeff Barr @ AWS asking if they could integrate its lifecycle into AWS' IAM SSL ecosystem.
        userbinator 11y ago
        I think servers are patched a bit more often than 90 days now, so the expiration period is not that short.
        toomuchtodo 11y ago
        Patching servers and cycling SSL certificates are apples and oranges. Depending on the size of the org, you may have several divisions that need to have new key/certs distributed to them, and their environments updated through change requests processes. Not everyone are teams of 5-20 on EC2 or DigitalOcean.
- jepler 11y ago
  He has posted on at least one other occasion on mozilla's bugzilla, see this 2010 comment on another certificate-related bug: https://bugzilla.mozilla.org/show_bug.cgi?id=556468#c33 https://bugzilla.mozilla.org/show_bug.cgi?id=556468#c33
  - jepler 11y ago
    And also on this bugzilla item, again questioning whether Mozilla is properly following procedure WRT letsencrypt https://bugzilla.mozilla.org/show_bug.cgi?id=1204656 https://bugzilla.mozilla.org/show_bug.cgi?id=1204656
    - onli 11y ago
      This one I saw. I did not count it as mozilla involvement. Nonetheless good to have it linked here to see the train of thought.

jpdus 11y ago

From the comments: A PITRA (point-in-time readiness assessment) is sufficient to begin issuing if the formal audit follows in a timely fashion. It's not possible to have a formal audit of the issuance process unless you are issuing; requiring it to begin issuance would lead to a chicken-and-egg problem. Gerv

rgbrenner 11y ago
not really true. Let's encrypt is cross signed with IdenTrust. So whether the ISRG certificate is trusted or not, they can still issue valid certificates.

idibidiart 11y ago

Iran gamed the CA system (DigiNotar?) to spy on their citizen's https traffic. http://www.theregister.co.uk/2011/09/06/diginotar_audit_damning_fail/ http://www.theregister.co.uk/2011/09/06/diginotar_audit_damn... So it's OK to be paranoid, IMO. The CA system is radically flawed in that it can be gamed by state actors.

rockdoe 11y ago
Completely and utterly irrelevant to the issue at hand.
- idibidiart 11y ago
  The issue at hand matters far less than the issue that the issue at hand is part of :) which is the CA system itself. You can try and move the corporatist agenda for a two tier security system where the mighty and powerful have all seeing eyes and the consumer is told that all their https communication is secure and private.

idibidiart 11y ago

The issue at hand matters far less than the issue that the issue at hand is part of, which is the CA system itself. With the CA system, you have a two tier threat model where the mighty and powerful can have all seeing eyes and the consumer is told that all their https communication is secure and private. Certificate Transparency and other proposals may help fix this, but the first thing to do is to acknowledge that there is a problem, and so far the major players have yet to acknowledge the fundamental problems of the current CA system. LetsEncrypt makes it worse by promoting the wider adoption of this flawed system. Certainly won't help keep us secure against the Chinese or anyone with big resource. Update: Stating that there is no solution without any knowledge of the existing solutions is itself part of the problem. Let's start with this: https://github.com/okTurtles/dnschain https://github.com/okTurtles/dnschain

userbinator 11y ago
What alternatives are there? SSH-style "you have not visited this site before, do you want to trust it?"? Distributed trust is a hard problem.
- idibidiart 11y ago
  Stating that there is no solution without any knowledge of the existing solutions is itself part of the problem. Let's start with this: https://github.com/okTurtles/dnschain https://github.com/okTurtles/dnschain

idibidiart 11y ago