• 0 Posts
  • 47 Comments
Joined 1 year ago
cake
Cake day: June 29th, 2023

help-circle







  • I actually agree. For the majority of sites and/or use cases, it probably is sufficient.

    Explaining properly why LE is generally problematic, takes considerable depth of information, that I’m just not able to relay easily right now. But consider this:

    LE is mostly a convenience. They save an operator $1 per month per certificate. For everyone with hosting costs beyond $1000, this is laughable savings. People who take TLS seriously often have more demands than “padlock in the browser UI”. If a free service decides they no longer want to use OCSP, that’s an annoying disruption that was entirely not worth the $1 https://www.abetterinternet.org/post/replacing-ocsp-with-crls/

    LE has no SLA. You have no guarantee to be able to ever renew your certificate again. A risk not anyone should take.

    Who is paying for LE? If you’re not paying, how can you rely on the service to exist tomorrow?

    It’s not too long ago that people said “only some sites need HTTPS, HTTP is fine for most”. It never was, and people should not build anything relevant on “free” security today either.


  • People who have actually relevant use cases with the need for a reliable partner would never use LE. It’s a gimmick for hobbyists and people who suck at their job.

    If you have never revoked a certificate, you don’t really know what you’re doing. If you have never run into rate-limiting issues with LE that block a rollout, you don’t know what you’re doing.

    LE works until it doesn’t, and then it’s like every other free service on the internet: no guarantees If your setup relies on the goodwill of a single entity handing out shit for free, it’s not a robust setup. If you rely on that entity to keep an OCSP responder alive for free so all your consumers can verify the validity of your certificate, that’s not great. And people do this to save their company $1 a month for the real thing? Even running the shitty certbot in compute has a larger cost. People are so blindly in love with this “free” garbage. The fanboys will never die off



  • I can’t answer this with confidence, but I was thinking the link in the email opened in the default browser, which wasn’t Tor in their case. Or something in the email client perhaps. Ultimately, I have no idea what happened and I was just speculating





  • There are many ways your real IP can leak, even if you are currently using Tor somehow. If I control the DNS infrastructure of a domain, I can create an arbitrary name in that domain. Like artemis.phishinsite.org, nobody in the world will know that this name exists, the DNS service has never seen a query asking for the IP of that name. Now I send you any link including that domain. You click the link and your OS will query that name through it’s network stack. If your network stack is not configured to handle DNS anonymously, this query will leak your real IP, or that of your DNS resolver, which might be your ISP.

    Going further, don’t deliver an A record on that name. Only deliver a AAAA to force the client down an IPv6 path, revealing a potentially local address.

    Just some thoughts. Not sure any of this was applicable to the case.

    There are many ways to set up something that could lead to information leakage and people are rarely prepared for it.





  • Numbers give the wrong impression that one version follows another. Debian release channels exit alongside each other individually. Giving the release channels names helps to make that distinction. It also makes for an easy layout of packages in APT repositories.

    Sid is and always has been Sid. If you were to assign numbers, what number should replace that name? There are perfectly working labels for release channels and there is no reasonable replacement.