[ DNS & domains ]

Authoritative and recursive DNS answers compared

Distinguish a zone’s authoritative answer from a resolver’s answer so DNS disagreements lead to the right authentication investigation.

By the EmailAuth research team · 6 min read

[ Key points ]

  • A direct authoritative observation and a recursive observation answer different questions.
  • RD and RA do not demonstrate that a response was freshly resolved.
  • Consistent names, endpoints, times, flags and transport make comparisons useful.
On this page

Ask who produced the answer

Two DNS lookups can disagree without either tool being broken. One may ask an authoritative server for the zone's current data; another may ask a recursive resolver that holds an earlier answer. A useful authentication investigation records which path produced each observation instead of treating a green check from any lookup site as the final state of the domain.

An authoritative server answers from the zones it serves. A recursive resolver works on a client's behalf, consulting other servers or reusing cached information. A stub resolver generally passes questions to a recursive service. The same software can perform more than one role, so identify the role of this particular response rather than relying on a product name. RFC 1034, sections 4.3.1 and 5.3.1.

This distinction matters when a newly published DKIM selector appears in the DNS host's tools but remains unavailable to a verifier. The authority can establish what a server currently serves; the recursive result helps explain what one receiving path may see. Neither observation stands in for the other.

Establish the delegation before choosing an endpoint

Start with the exact owner and type implicated by the message or service instruction. Follow the delegation to the zone responsible for that owner. A child zone can have different authoritative servers from the parent, and a provider alias can lead into a different zone again. The DNS host displayed on an old ticket may no longer be the delegated host. RFC 1034, sections 4.2.1–4.2.2.

Prepare a list of authoritative server names and their addresses, then query the relevant servers directly. Preserve which address was contacted. If different authorities return different records, investigate the DNS publication or zone-serving problem first. A resolver can legitimately encounter any available authority; choosing the server with the desired answer does not resolve the inconsistency.

A trace can help inspect the delegation path, but it is a separate observation. It does not reproduce a receiver's cached path, forwarding configuration, source network or DNSSEC validation behavior. Label a trace as a trace rather than presenting it as the mail provider's query history.

Interpret flags with their limits

The AA flag identifies an authoritative answer in the DNS response. RD records that recursion was requested; RA advertises recursion availability. RD and RA do not prove that the server performed a fresh lookup for this response. A recursive server can satisfy a request from cache. RFC 1035, section 4.1.1.

For a direct authority check, request no recursion and inspect the actual reply. A response without AA, a referral, REFUSED or a timeout calls for interpretation; the server address alone does not make every reply authoritative. Where aliases appear, keep each owner and target visible and identify which zone supplies the final data. Authority over the alias owner does not automatically establish authority over its target.

Retain the complete question, answer and authority sections. A command that prints only TXT strings hides context needed to distinguish old cached data, a referral, an alias and an absence response. Short output is useful for quick reading after the full diagnostic record has been saved.

Compare a fictional pair of resolver observations

Imagine a fictional selector blue._domainkey.example.com. At 10:00 UTC, two relevant authoritative servers return the newly intended public-key record. At 10:02, Resolver A returns an older key with 420 seconds remaining, while Resolver B returns the new key. These times, results and TTLs are invented to illustrate reasoning; no network measurements were made.

That pattern is consistent with different cache histories, but the observation alone does not prove the cause. Check that all questions used the same name, type and class, that the server endpoints were recorded, and that no split view or forwarding policy changed the answer. If Resolver A later refreshes from an inconsistent authority, simply waiting may not solve the problem.

The proposed incident note would state: “Authorities sampled agree on the new key; Resolver A still returns the old key; receiver outcome not yet observed.” It would not state that every receiver is updated or that authentication now passes. Keep the older public key's availability and the sender's signing transition under a coordinated rotation plan rather than deleting records to make a lookup screen look tidy.

Separate data problems from reachability problems

A recursive SERVFAIL does not establish that a record is absent. An authority may be unreachable, a delegation may fail, or DNSSEC validation may prevent the resolver from accepting the answer. A direct answer from one server narrows the investigation but does not prove that all dependencies are healthy for the recursive resolver.

Transport also matters. DNS requires TCP support, and a truncated UDP response can require a TCP retry. Compare a proposed dig @AUTH_IP blue._domainkey.example.com. TXT +norecurse with the same query using +tcp when transport is implicated. Do not alter firewall or security settings merely to make the test pass; have the responsible operator review the diagnosed path. RFC 7766, sections 5–6.

Make the comparison reproducible

Use one worksheet row per query: UTC timestamp, operator location, server endpoint, recursion choice, owner, type, class, response code, flags, records, TTLs and transport. Repeat selected observations at a documented interval and preserve the earlier rows. DNS caches evolve, so overwriting the first result removes useful timing evidence.

Finish with a controlled message through the affected sender and retain the identities and trusted receiver results. DNS comparison can explain a dependency; it cannot demonstrate inbox placement or a particular receiver's internal lookup. This article supplies a proposed workflow and a fictional comparison. Original timestamped authoritative and recursive answers remain required before its catalogue evidence gate can be closed.

[ Takeaway ]

Identify the server role and complete query context before interpreting two different DNS answers.

[ By role ]

What this means for you

Choose your role for practical next steps.

DNS operator

Separate inconsistent authority from caching

A resolver can only retrieve the data and reachability offered by the DNS path it encounters.

Next steps

  1. Compare relevant authoritative endpoints with recursion disabled.
  2. Escalate differing zone answers before declaring the resolver stale.

Messaging support engineer

Retain the receiver-facing path

A public lookup site cannot reveal the affected receiver’s internal lookup history.

Next steps

  1. Record the exact selector and signing domain from the affected message.
  2. Label each recursive endpoint and preserve its complete result.

Incident coordinator

Write conclusions at the evidence’s scale

The incident needs a bounded statement about sampled paths, not a global propagation claim.

Next steps

  1. State which authorities and resolvers were checked and when.
  2. Keep DNS restoration and successful message verification as separate milestones.

[ FAQs ]

Frequently asked questions

Does a recursive answer always come from cache?

No. A resolver can use cached data or perform resolution. Its role and recursion flags do not by themselves reveal how recently it contacted authority.

What does AA tell me in a direct server response?

AA marks an authoritative answer. Inspect the actual response and alias context rather than assuming an endpoint is authoritative for every name you send it.

Does RA mean the resolver just fetched my DKIM key?

No. RA indicates recursion is available; RD records the request for recursion. Neither flag is a freshness receipt.

Why query more than one authoritative server?

Different servers may serve inconsistent data or differ in reachability. A recursive resolver may encounter another authority, so one desired answer does not resolve a disagreement.

Can dig +trace reproduce a mail receiver’s lookup?

No. A trace examines a delegation path from its own environment. It does not reconstruct the receiver’s cache, forwarding rules, network or validation behavior.

Why might the parent zone’s name servers be the wrong servers to query?

The requested owner may be inside a separately delegated child zone. An alias can also take the lookup to a target in another zone.

What does a direct positive answer plus recursive SERVFAIL prove?

It shows that the sampled authority supplied data while that recursive path failed. It does not identify a unique cause or establish that validation and every other authority are healthy.

When is a TCP comparison useful?

When truncation or transport reachability is implicated. Repeat the same name, type and server with TCP and retain both results; TCP support is part of DNS requirements.

[ Sources ]

Sources

  1. RFC 1034: Domain names: concepts and facilities
  2. RFC 1035: Domain names: implementation and specification
  3. RFC 7766: DNS Transport over TCP: Implementation Requirements

[ Glossary ]

Glossary

Authoritative server
A DNS server serving data for a zone for which it has authority.
Recursive resolver
A service that resolves a client’s question by consulting other servers or using cached information.
Stub resolver
A resolver component that normally relies on a recursive service to perform resolution on its behalf.
AA flag
The DNS response flag identifying an authoritative answer, interpreted in the context of the question and answer.
Referral
A DNS response directing resolution toward servers responsible for a delegated zone rather than supplying the requested final data.

Enterprise provenance

Why top financial institutions and governments retired their email security dashboards.

Traditional email authentication platforms sell you a dashboard and ask your engineers to do the labour. In high-stakes environments, manual oversight means delayed enforcement, configuration drift and unmonitored shadow domains.

EmailAuth is the agentic email provenance platform. Our agents enforce p=reject, keep third-party senders aligned and move your domains to full RFC 9989 compliance, the new DMARC standard,, with human-in-the-loop auditability.

Governments Large enterprises Tier-1 financial institutions

SaaS · Dedicated tenancy · Air-gapped on-prem · BYOK encryption

Knowledge Base