47 karma · joined September 30, 2023
I've been looking at contrib.resolver in urllib3.future and noticed that you only parse A,AAAA,HTTPS from responses, but you do not handle a CNAME. I wonder, because kubernetes service can return a CNAME for app to follow:
https://kubernetes.io/docs/concepts/services-networking/serv...
Just wonder, if this is something worth pursuing, to follow one CNAME...
I'm fully aware this is not a recursive resolver, but just wondering about a CNAME response from coredns here...
I was expecting a pcap library that would try to catch server side certificates from tls handshake and block them with nflink or something, however not sure if serverhello is encrypted or not.
Recursive servers do not need public root servers because you could run a root yourself at 127.0.0.99 and point recursive to it, or bootstrap recursive one with this data. Of course, you need full control of the recursive server to set it up this way.
Taking down all public root servers (all 1723 of them, according to public data) may have no impact on a properly set up recursive servers, because root zone is public and you can host it yourself.
We use dot "." to split labels. wwww.example.com has 4 labels:
"www" . "example" . "com" . "" (NULL, root label, 0 length).
Because we can't print NULL/"", it looks like a single dot at the end. But this single dot separate com from root label, which is empty "". We ommit this label and this dot.
Every other record you mentioned never gained any adoption, simply because this "database" is meant to be used by programs, not humans. Sure, I may use SOA email to contact a DC, but this is exception, not a rule.
Therefore, from the perspective of a person that does DNS courses for technical staff, I conclude that average Joe will never be able to set TXT records properly, and professionals will not do it either because of the reasons I mentioned.
And while I encourage you to pursue your goal, adoption willy be close to none so I'm not so worried about that. Adoption of this will be much lover that adoption of HWINFO or LOC records (ignore RFC 8482, I mean user provided HWINFO data). Your proposal will be used by 50 people total, optimistically.
For all use cases, we have http invented already. Storage for http is also a solved problem.
Network without DNS is un-useable. Apologies for my english, it is not my native language.
Another example, lets say you spin up a QuicTun, nacl-based encapsulation tunel, encrypted, running over IP + UDP as a layer7 process, in userland, there is no kernel implementation. For me, this vpn is Layer2 (when tap mode) or Layer3 (when tun mode) regardless that it uses all 7 layers to establish the tunnel. Sure it uses all layers to establish the connection but once it is done, I have layer3 device for further use.
Same for DNS - for me it is Layer3 because of the crucial function it provides, so I could rather say it is functionally on the same layer as IP in a sense, it is crucial to make a connection. Does not matter that it uses all 7 layers to get me the network (A)ddress / (AAAA)ddress, it is still functionally a network layer for me. Only once this is establised, I can exchange data.
And while DNS is running on layer7 - for me personally, the function it performs is "to allow connect() to a string as opposed to a network address (IP/IPv6)" and because of that, it is a Layer3 for me on the functional level, same as a VPN software to access some corporate stuff. Does not matter that VPN uses TLS and DNS under the hood and span across all 7 layers to encapsulate my data, functionally it is layer3 for me.
I do a lot of corporate DNS trainings and people are surprised about this view. Software developers are unhappy that we start the "DNS course" with IP basics, unicast, anycast, 101 overview of BGP, ARIN, provider independed vs Provider Aggregatable address space. Developers would rather learn about DNS, not what an IP address is. Network guys sits happy on the course, but they are not happy to hear my view that:
"there is no DNS without network and there is no network without a DNS"
However, slowly but surely once the full training is finished, both sides understand that IP/IPv6 and DNS are in a symbiosis and can't be separated. We touch a lot of topics, including lack of interoperatibilty between IPv4 and IPV6, DNS64, XLAT464, etc.
Sure, if you run an air-gapped network in some military facility, you can get things working without DNS, but 99.999% networks can't work without a DNS.
So the way things work today is this, the way I see it:
1. We have a physical layer -> a wire
2. We have a data link layer -> ethernet + ARP or NDP are common here
3. we have a network layer -> currently, two networks are widely used: IPv4 and IPv6
4. We have a transport layer: tcp and udp and quic
7. We have an actual data, encrypted or not.
So where the DNS fits in this model? People say it is build on top of UDP, so it must be layer7, right? But because of the core function of the DNS protocol, I would put DNS into Layer3, the same as I would put ARP and NDP on layer2 even if NDP is built into IPv6. I would put it on Layer 2 solely by the function it provides.
A sole purpose of the DNS is to make a connect() possible without using direct network addressess, but client use "a string" as a destination. For me this is a Layer 3 protocol, because it is a helper protocol used to establish a connection.
Once you have a connection, on top of it you use other layers to transfer data reliably.
Your solution puts a Layer 7 data into a Layer 3 component by using TXT records. And if there were no TXT records invented, you could still encode binary data in the AAAA records, 8 bytes at a time and hack it to store arbitrary data that a custom "client" can process in 8 byte chunks.
And while clever, this is not the right tool to store layer 7 data.
It is like sitting on an island with a hammer. You need to cut a tree and says: well, I can use hammer to cut the tree or I can try with bare hands. What should I do? While I agree hammer has a higher chance of succeeding, I would advice to find a sharp rock instead. Find or invent a better tool for the job.
First, because of a distributed nature of the DNS, clients do not directly fetch this data. Instead, resolvers and DNS forwarders do the work and cache it.
Secondly, because displaying or converting this data to html requires a special website or a dedicated client, at which point this dedicated component should use a different storage than DNS. Whatever that may be, p2p, etc.
Third, as the impact on the DNS ecosystem is problematic here. I will say something controversial by stating this: DNS had been designed in one goal, for you or your program to connect "to a string" that you or a string owner control and manage, as opose to connecting to a network address (IP) that you usually do not own, that can change at any time, and that it is meaningless and hard to remember. Over the years, this basic principle extend a bit with introduction of some policies being stored in the DNS that is meant to help the domain owner setting up some boundaries or policies, like SPF or CAA records.
Storing website data is not something I would recommend. Again, you can argue that this "client" only prints TXT data to users that would not normally be able to access it (average internet browser user) but clearly it goes far beyond this, encouraging storing domain content (data) into a DNS.
Fourth, security. Very few domains implements dnssec or curvedns and clients do not use DoH en masse. A mitm is a serious threat to the client rendering this data.
Fifth, if this gains wide adoption (doubt it), resolvers will stop caching TXT records as a policy to save resources for crucial data, and the crucial data is A(dress) record(s).
Because the purpose of the DNS on a high level view, is too make a connection between two programs or users. I know that's a simplification, but this is the core idea. Sure, it has reverse DNS mapings to assing a string onto a network address, but 99.99% of the time DNS job is to make connect() possible. So you connect() to Amazon using DNS, Amazon contacts payment processor using DNS, they connect() to Visa which connect() to your Bank and every connection in this payment processing chain use DNS to maka it possible and reliable.
However, I'm a fan of geodns concept in general, and we have ECS standard as well to help with better answers. This allows small players or individuals to do what Big companies do, but without running anycast network, which is very expensive at every step of the journey: from getting ASN, then PI address space, paying yearly fees, then getting bgp peering and operating it efficiently. Thousands of $ per month and a required expertise. Instead of just spinning up 5 cheap vps and put a smart DNS in front, accepting all the tradeoffs, for $50 per month.
I do agree however, that some of the operational cost (marginal) is shifted to other parties because of small TTL need for this to work sensibly. But still, majority of traffic goes to my authoritative DNS servers that I operate, as the higher zones have a big TTL anyway. You can't geodns a glue record, it is static and it is cached properly.
So who bear this cost? End user pays for the bandwidth so an extra packets for lookup is not an issue. ISP resolvers? A bit yes, but they can enforce a policy on minimal TTL if this is too expensive for the ISP... Clearly it is not. At the end of the day, customer (ISP customer) will pay the bill if ISP needs to spin up more resolvers.
However, the problem was a missing synchronisation of readed entries between your home and office PC. Mobile device was not a thing back then but slowly started main stream adoption. I mean, very few had smartphones.
Google reader was revolutionary, solving this exact problem for free and mobile app quickly followed. It caused almost all offline RSS readers to decay and starve to death.
And then few years passed by, and this is the moment they pull the rug, but the damage was done. Luckily, Feedly and others quickly emerged, and smartphones were a thing. Also, some self hosted RSS readers came to existence and get polished.
But how ventoy achieve systems booting and detecting the iso programmatically? I know that initially, there seems to be grub2 reading usb file list and then it reads iso eltorito boot image and boots from it, but how exactly the OS itself know about the virtual drive? Is there an interface in uefi or bios that allow emulate optical drive, or some other standard is at play and operating systems just happens to support it? Window 10/11 pe / installer, for example, reports the iso in diskpart as virtual disk. Not an optical drive. Windows kernel actually mounts the iso somehow automatically.
Can someone point me on how this actually works, please.
I'm an old it guy, without rss I will not bother checking the site ...
This "elitism" judgement will do no good in the long run, just saying.
Would it be possible to make computer keyboard somehow customizable and keep it as a permanent cookie value or local storage?
I'm not a fan of nodata responses with empty authority, same as nxdomain responses.
Missing TCP listener. While not required by any RFC, some ccTLD Registries enforce udp+tcp before delegating, like .de
Or CNAME rrdata. Or MX rrdata. Albeit, there is little point in using it this way, as oppose to just add A record in your own domain anyway...
For people without a domain and a static IP, this may be useful too too publish a page, and possibly obtaining ACME tls cert.
See RFC 2308.