4 ms·
"The assertion of this injunction is, in essence, that if there is any technical possibility of denying access to content by a specific party or mechanism, then
by 1vuio0pswjnm7 5y ago
"The assertion of this injunction is, in essence, that if there is any technical possibility of denying access to content by a specific party or mechanism, then it is required by law that blocking take place on demand, regardless of the cost or likelihood of success."
Technically, this block does not stop anyone from getting the IP address for a domain. Anyone can resolve a domain name using public information that is disseminated from domain name registries, domain name registrars and other authoritative DNS providers. Quad9 is just a third party DNS provider, not an authoritative source for IP addresses. In theory, third party DNS providers could refuse to provide (resolve) the IP address for any domain name. They could do this on their own accord, to suit their own interests, or at the behest of anyone, e.g., an end user, a financial contributor (donor), an interested corporate partner, or perhaps pursuant to a court-ordered injunction.
In fact, this in exactly what Quad9 does: they block domains. They advertise this capability on their website, where even the most non-techical reader could find it. From the "About" page:
"Quad9 blocks against known malicious domains, preventing your computers and IoT devices from connecting to malware or phishing sites."
Third party DNS has a number of potential problems; filtering is one. Funny how people have literaly turned that problem into a selling point. For example, OpenDNS, now part of Cisco, started a business doing DNS-based filtering.
Personally I fail to see why third party DNS (ISP-provided DNS or so-called "open resolvers") remains a preferred method of retrieving IP addresses or other RRs. IMO, there is no technical advantange anymore.
Many years ago I wrote a system for resolving domains without using recursion, using only authoritative queries, never setting the RD bit. It was very fast. Faster than a cold cache, IME. It could actually get faster as it acquires more addresses of authoritative servers, because it does not need to look them up again. It "learns". The best aspect though is that there are no unecessary third party middlemen. Third party DNS providers are not authoritative sources for any RR. They are middlemen. They do not operate for free. They are potentially subject to influence from whomever pays the bills.
People often discuss "privacy" when they discuss third party DNS service. IMO, using a shared third party DNS cache seems antithetical to "privacy" (not to mention "security"). In any event, it enables filtering by someone who is not an authority for the DNS data they are serving, a middleman. This is the view of an end user, not a corporation nor a developer working for one.
- zinekeller 5y ago> IMO, there is no technical advantange anymore. So, have you tired resolving a domain on a high-latency connection? Or live in an area where the internet routing don't make sense but you don't have a choice (or worse, you have a choice to different ISPs who all have different weird routing)? Not everyone has a good connection, and this is doubly true for residential connections. In those cases, "smart" DNS providers can steer the users to the fastest route, which in some cases resolves to a better server than the answer given to when a user does a direct recursive answer (because the authoritative servers directs the user into a "bumpy" routing). That's the (original) selling point of 1.1.1.1: they will use Cloudflare's knowledge of your routing to give you a better server (even if the domain you're asking for is not from Cloudflare's network).
- 1vuio0pswjnm7 5y agoYes. Everything I do is designed to deal with those conditions. I never assume a powerful computer or a reliable connection. But nothing is ever "once-size-fits-all". There could be situation where, e.g., a website's authoritative DNS servers are barely functional, but a cache like Cloudflare's has a copy of the RRs. IME, that is quite rare. I store the RRs permanently in a custom zone files once I retrieve them. Then query the data from a loopback-bound authoritative DNS server. No subsequent queries for those RRs leave the network interface. This is faster than 1.1.1.1 .
- zinekeller 5y agoYes, I do get that your system caches the authoritative servers, but as I said above there are times where the connection is "weird" and some DNS revolvers are better-equipped to solve them. Some authoritative DNS administrators wrongly used (for example) MaxMind GeoIP (don't do this, use ASes to differentiate connections when it comes to DNS steering) to steer their DNS requests, and it ends up that the user is getting a suboptimal server. Here's a real-life example I encountered: Wikimedia (which operates Wikipedia et al.) has historically routed Japanese connections to Singapore, where they're geographically close, but due to how the internet backbone in the area works (it goes via Hong Kong, where switching time adds up to the latency) it made more sense to route them to WM's Los Angeles servers (and some ISPs have done this before WM has formally rerouted it to Los Angeles). Or even a more frustrating one: I have routed SingTel users to Hong Kong despite also operating a server to Singapore because their routing is so bonkers that SingTel will route the connection to America and back just so they can access a server that they can physically go by using the metro (yes, I have contacted their NOC. No, they haven't changed anything). Will this break DNSSEC? Absolutely. Will ordinary users care? While I'll care about correctness, more users will care about the speed and quality of their connection rather than correctly routing to a suboptimal server.