services/New-DNS.md
... ...
@@ -10,11 +10,14 @@ If running your own resolver is not possible or undesirable, you can choose one
10 10
11 11
You can also use the globally anycasted a.recursive-servers.dn42 but you won't have any control over which instance you get. This is a **very bad idea** from a security standpoint.
12 12
13
-# *.recursive-servers.dn42
13
+# Instances
14
+The new DNS system has three different components: *.recursive-servers.dn42 and local resolvers responsible for handling queries from clients and validating DNSSEC. *.delegation-servers.dn42 and *.master.delegation-servers.dn42 are a normal master-slave setup for providing the few official infrastructural zones.
15
+
16
+## *.recursive-servers.dn42
14 17
These are simple resolvers capable of resolving dn42 domains. Every operator gets a single letter name pointing to addresses assigned from his own address space and is strongly encouraged to use anycasting across multiple nodes to improve reliability. There is also the global anycast a.recursive-servers.dn42 which includes some/all other instances.
15 18
16
-# *.delegation-servers.dn42
19
+## *.delegation-servers.dn42
17 20
These are simple authoritative servers for the dn42 zone, rDNS and a few DNS infrastruture zones. Every operator gets a single letter name pointing to addresses assigned from his own address space and is strongly encouraged to use anycasting across multiple nodes to improve reliability. There is no anycast instance because that would make debugging much harder and *.recursive-servers.dn42 instances should do loadbalancing/failover across all instances listed in the registry.
18 21
19
-# *.master.delegation-servers.dn42
22
+## *.master.delegation-servers.dn42
20 23
These instances do not serve any clients. They poll the registry regularly and rebuild and resign (DNSSEC) the zones as needed. If any zone changes, all *.delegation-servers.dn42 instances are notified ([RFC1996](https://tools.ietf.org/html/rfc1996)) which then load the new zone data over AXFR ([RFC5936](https://tools.ietf.org/html/rfc5936)). The pool of masters is intentionally kept very small because of its much higher coordination needs and also the lacking support of a multi-master mode in many authoritative server implementations. The masters are only reachable over dedicated IPv6 assignments which are set up in a way that any master operator can hijack the address of a problematic master without having to wait for its operator to fix something.
... ...
\ No newline at end of file