title: "Guidance on DNS Message Composition in Constrained Networks" abbrev: "DNS in Constrained Networks" category: bcp
docname: draft-lenders-dns-cns-latest number: date: consensus: true v: 3 submissiontype: IETF area: "Applications" workgroup: TBD keyword:
- Internet-Draft
- CBOR
- DNS venue: group: TBD type: Working Group mail: [email protected] arch: "nicfs.nic.ddn.mil:~/namedroppers/*.Z" github: "anr-bmbf-pivot/draft-lenders-dns-cns" latest: "https://anr-bmbf-pivot.github.io/draft-lenders-dns-cns/draft-lenders-dns-cns.html"
author:
- fullname: Martine Sophie Lenders organization: Freie Universität Berlin abbrev: FU Berlin email: [email protected]
- fullname: Thomas C. Schmidt organization: HAW Hamburg email: [email protected]
- fullname: Matthias Wählisch organization: Freie Universität Berlin abbrev: FU Berlin email: [email protected]
normative: RFC1035: dns
informative: RFC4944: 6lowpan RFC7228: cns RFC7858: dot RFC8484: doh RFC8724: schc RFC8499: dnsterm RFC6763: dnssd I-D.ietf-core-dns-over-coap: doc
--- abstract
This document provides guidance on the composition of DNS messages in constrained networks, where the link layer may restrict payload sizes significantly and batteries challenge power consumption.
--- middle
Many IoT scenarios rely on constrained nodes {{-cns}} and need the Domain Name System (DNS). Constrained nodes {{-cns}}, however, challenge DNS resolution for two reasons. First, IoT networks, such as IEEE 802.15.4 and LoRaWAN, often limit the payload of the data link layer significantly in terms of size, which prevents common (larger) DNS responses. Second, constrained nodes are often battery powered and follow a strict duty cycle, which would benefit from minimal number of DNS messages to reduce unnecessary device wake-ups. Adoption layers such as 6LoWPAN {{-6lowpan}} and SCHC {{-schc}} provide fragmentation and compression to overcome the problem, but do not help in principle. Fragmentation requires more buffer space to account for lost fragments in lossy networks, and compression introduces additional processing overhead.
This document provides best common practices on DNS behavior, to reduce fragmentation and power consumption in constrained networks when IoT nodes resolve names.
A "DNS server" is a server that provides DNS information to a querying DNS client. For the purpose of this document, server and client may communicate based on any DNS transport, not just DNS over UDP {{-dns}}, but also DNS over TLS {{-dot}}, DNS over HTTPS {{-doh}}, or DNS over CoAP {{-doc}}.
The terms "constrained node" and "constrained network" are used as defined in {{-cns}}.
{::boilerplate bcp14-tagged}
Nodes within a constrained network that implement DNS are assumed to be stub resolvers {{-dnsterm}}. This means they only query information from a recursive DNS resolver and they MUST NOT distribute any DNS information received from an upstream DNS server.
A DNS server that is aware that the querying node is a node within a constrained network SHOULD resolve a CNAME or PTR record until the resource record type originally requested by the node is reached. This reduces the number of message exchanges within a constrained network.
The DNS server SHOULD send compact answers, i.e., omit additional or authority sections in a DNS reply. Additional and authority sections should only be included if they help a DNS client to reduce queries. One such example is DNS-SD {{-dnssd}}, where the answer does not only need to include an SRV record but also TXT and A/AAAA records to make decent use of the reply.
TBD: Provide more specific example use case? Should the list be almost complete?
In the case when DNS clients act as DNS servers, resolving CNAME and PTR records at the upstream DNS server may lead to incorrect DNS information forwarded by the client. As such, this document prohibits the distribution of such information in {{sec:node-considerations}}
TODO more security
This document has no IANA actions.
--- back
{:unnumbered}
TODO acknowledge.
- Carsten Bormann
- Ben Schwartz