this post was submitted on 19 Apr 2024
28 points (83.3% liked)
Privacy
31954 readers
447 users here now
A place to discuss privacy and freedom in the digital world.
Privacy has become a very important issue in modern society, with companies and governments constantly abusing their power, more and more people are waking up to the importance of digital privacy.
In this community everyone is welcome to post links and discuss topics related to privacy.
Some Rules
- Posting a link to a website containing tracking isn't great, if contents of the website are behind a paywall maybe copy them into the post
- Don't promote proprietary software
- Try to keep things on topic
- If you have a question, please try searching for previous discussions, maybe it has already been answered
- Reposts are fine, but should have at least a couple of weeks in between so that the post can reach a new audience
- Be nice :)
Related communities
Chat rooms
-
[Matrix/Element]Dead
much thanks to @gary_host_laptop for the logo design :)
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Because:
If you query their "unsecure" servers, it works:
I would never use a provider that has the means in place / likes to filter the DNS - you never know when a govt will ask them to kill a certain domain (even for the "unsecure" servers) or redirect things.
Don't all providers have the ability to filter things?
Personally I have very few problems with DNS providers filtering out malware, that sounds like a benefit to me.
It depends on what you can consider "the ability". If by ability you mean have to deploy a team of engineers working for a week to make it happens, that's okay, if they've their system built for it things are different.
I can do this in like 5 seconds with my PiHole and not only am I not a network engineer, I would encourage people to never employ me as such.
So for an actual business that has a bigger budget than me ($0) and more hours to devote to it than me (.02), shouldn't it be less of a problem?
Exactly and consider Cloudflare for instance, adding an “if domain block” is easy but then once you’ve thousands of servers running the same piece of software across the globe deploying updates and features becomes way slower and way harder. You’ve to consider tests, regressions, a way to properly store and sincronize the blocklists across nodes etc…
I'm not saying it can be done, because it can. But it will take longer and it will be a problem for someone. Besides you only have that point and click interface in your PiHole that allows you to do it in .02 because someone spend a few hours developing the feature. :)
This is what we're trying to explain to you, this is how DNS works. Those thousands of servers? Recusrive DNS resolvers, ran by Cloudflare. All watching and caching the records from Cloudflare's authoritative nameservers in near real time, because that's how it was designed. You don't need to test for regressions, figure out how to properly store and synchronize the "blocklist" (it's not a blocklist, it's changing a domain record or simply using a CNAME to point to the registrar) or whatever else, because DNS is continuous, and it was designed to do what you're describing, in the 90's.
Yes, if you're updating your infrastructure, you'd want to test. But this isn't that.
Ever ran into an expired domain and thought about how the registrar can just park an expired domain and make it an ad for themselves? That's just them adding a CNAME in their authoritative nameservers, which gets distributed globally. The prior delinquent owner can still be hosting, but because they don't have the authoritative nameserver they can't use the domain anymore.
The ability to selectively respond to DNS requests is integral to the function of DNS. The only real issue here is that there isn't a standard response code indicating the reason for not returning the record like there is in http
The availability of such feature and how useful it might be to block something is dependent on the actual implementation (software) you're using.
Making a DNS server not respond to queries for a specific name is trivial for any DNS provider to implement, this is not a situation where they would have to develop months worth of new features to support that if the government asked.
It might not be that easy, you're thinking about one single server running some kind of DNS server you're familiar with. When we're talking about Quad9, Cloudflare etc. were talking about hundreds of servers across the planet, highly distributed solutions that rely on multicast and other non-trivial techniques. If you've to change a system like that to add the ability to block something, trust me, it won't take a few hours.
DNS is literally distributed by design. It's how it works. Even if a deployment was done for some reason, it would not take more than a single engineer (an engineer really isn't even necessary for this, because again, it's built into DNS).
You know you can setup a single instance of PDNS or other resolver and by default they all work in a non-distributed way. You assuming to much, and again while it is likely that most providers running custom stacks already have the functionally built in, it isn't a guarantee and once you've thousands of servers running the same piece of software across the globe deploying updates and features becomes way slower and way harder.
That is not what they do, though. Just because a non standard configuration is possible doesn't mean that's the best thing to use. DNS, by design, uses authoritative nameservers, which is what cloudflare and quad9 host. These authoritative hosts distribute their records to caches (usually just recursive DNS resolvers) to ease and distribute the load. It's literally in all of their documentation, and explained in pretty plain english on their pages.
https://www.cloudflare.com/learning/dns/what-is-dns/ https://www.quad9.net/about/
When a record is updated in your domain (or cloud) provider, it is distributed via an authoritative nameserver hosted by that company. These get distributed to the root name servers, which then distribute the records to other authoritative nameservers.
I don't know why you're arguing over this, when it's one of the first things you learn in information systems and networking. Sure, there's a lot of stuff for the infrastructure. But the way DNS works on these hosts is still the same, and blocking a single record is not difficult and does not take extra engineering effort. The authoratative hosts simply change their records and it's done. DNS takes care of the rest.
There's an entire wikipedia page on this: https://en.wikipedia.org/wiki/DNS_blocking
The thing is, everything you said is correct. But if you think they can just solve this globally for everyone and everything without delays by just pushing things their root servers or the first line of authoritative ones then what else can I say.
They can, because that's how DNS works. This is why when you update a record for your domain it's updated globally in near real time with multiple providers. I don't know how else to tell you that it already works this way. I work in the cloud, and deal with this stuff on a daily basis.
Depending on the TTL, right ?
Correct!
So, you know that "near real time" is different from actual real time.
You're clearly going keep nitpicking and changing the subject to things that don't matter and you're not willing to learn. Your misunderstanding of the fundamentals of DNS is no longer my issue.
Okay, so tell me something, in your "ideal world" replication-based only scenario, what happens if they're ordered to take down a specific A record that has a very large TTL, more than a govt would like?
It would literally be easier to add that capability to your own custom DNS server software. After all it is literally an "if query.name in blocklist then drop connection". Even replacing results would be simple as long as DNSSEC is not involved. You wouldn't have to add it though since all major DNS servers already include it because it is so simple and has legitimate uses, such as blocking malware control server names or ad blocking.
Yes, it is likely that most providers running custom generic or custom stacks already have the functionally built in and also yes, adding an "if" is easy but then once you’ve thousands of servers running the same piece of software across the globe deploying updates and features becomes way slower and way harder. You've to consider tests, regressions, a way to properly store and sincronize the blocklists across nodes etc...
How much simpler can I make this...
You have a primary 'master' server in the pool.
Replica/cache servers periodically ask the master for any updates.
Master gives a new update, which is a sinkhole for a marked malicious domain.
Replica/cache server now resolves malicious domain to the sinkhole address.
This is not a 'feature' you have to implement, it's a basic function of running a redundant DNS system.
I can do the same in something like adguard or pihole...
Governments don’t ask. They order. And it happens on a regular basis.
Yes, but if the provider doesn't have the capabilities baked in they'll take more time to comply or just not do it at all.
You really don’t want to ignore an order from a judge. And blocking websites is trivial.
Nothing is trivial at scale. When we’re talking about Quad9, Cloudflare etc. were talking about hundreds of servers across the planet, highly distributed solutions that rely on multicast and other non-trivial techniques. If you’ve to change a system like that to add the ability to block something, trust me, it won’t take a few hours and a LOT of testing will be required before pushing into production.
The ability to change address records at global scale is built into DNS. It's not a new thing.
Nothing is "built into DNS". DNS is a couple of RFCs that include specifications on how the thing should work. What features one implementation (software) has is decision of those who made it and nothing else.
What you said here is not really on topic, but it is literally part of DNS. I already explained it in my other comment, but here: