Skip to content
AdminKit

How do you read an SVCB or HTTPS DNS record?

SVCB and HTTPS records, defined in RFC 9460, publish connection hints in the DNS answer: which ALPN protocols a service speaks, a non-default port, or fallback IPs, so a client connects faster without guessing first. Clients that ignore them fall back to A/AAAA lookups. This AdminKit guide shows how to read one in dig output.

By AdminKit team, Netbase JSC · Updated

Key takeaways

  • An SVCB or HTTPS record tells a client how to connect to a service, not just where it is: which protocols it speaks (alpn), which port, address hints, and an Encrypted Client Hello key (ech).
  • Every record has three parts: a priority, a target name and a list of parameters. Priority 0 is AliasMode, a pointer to another name. Anything higher is ServiceMode, an actual endpoint with hints.
  • HTTPS is the web-specific version of SVCB. An AliasMode HTTPS record can alias a zone apex, which a CNAME can’t.
  • The records are additive. Clients that don’t understand them use A/AAAA as before, so keep those records in place.
  • AdminKit’s DNS lookup tool doesn’t query these types; use dig example.com HTTPS to check one.

What problem do SVCB and HTTPS records solve?

They let a client learn how to connect before it connects. As part of AdminKit’s DNS cluster, this guide picks up where the record types covered in how to read a DNS lookup result — A, AAAA, CNAME, MX, TXT, NS — leave off. Each of those answers a narrow question: an address, an alias, a mail destination, free-form text, or a zone’s authority. RFC 9460 defines a type that answers a different question entirely: not just “where is this service,” but “how should I actually connect to it.” SVCB (Service Binding) and HTTPS records let a zone publish, in the DNS answer itself, which application protocols a service speaks, which port it listens on, fallback addresses, and even TLS handshake material — hints a client would otherwise only discover after already opening a connection and negotiating.

HTTPS is a fixed specialization of the general SVCB record for web traffic specifically: a client resolving https://example.com queries the HTTPS type at that name. RFC 9460 §9 goes further and says clients must not use SVCB queries for https or http URLs at all, so a web server’s hints belong in an HTTPS record. SVCB stays available for other schemes that define their own use of it, DNS servers among them. Both share the same wire format defined in RFC 9460, so everything below about reading the fields applies to either type.

What are the three fields in an SVCB or HTTPS record?

Priority, target name and parameters, in that order. RFC 9460 §2.1 gives the zone-file form, and §9 restates it for HTTPS as Name TTL IN HTTPS SvcPriority TargetName SvcParams.

FieldExampleWhat it means
SvcPriority0, 1, 20 selects AliasMode. 1 and up select ServiceMode and rank endpoints; lower is preferred.
TargetNamepool.example.net. or .The host to use. In ServiceMode, . means the record’s own name.
SvcParamsalpn=h3,h2 port=8443Space-separated key=value hints. Ignored in AliasMode.

For a web service on the default port, the record sits at the hostname itself. A non-default port gets a prefix: hints for https://example.com:8443 live at _8443._https.example.com (RFC 9460 §9.1).

What is the difference between AliasMode and ServiceMode?

AliasMode points somewhere else; ServiceMode describes an endpoint. Every SVCB/HTTPS record starts with a 16-bit SvcPriority field, and its value changes what the rest of the record means:

  • Priority 0 — AliasMode. The record carries only a TargetName and no parameters. It tells the client “the service hints you want live at this other name, not here” — the client re-resolves SVCB/HTTPS at the target instead. This is conceptually similar to how a CNAME record redirects a lookup to another name, though it is its own distinct mechanism specific to this record type, not a CNAME.
  • Priority 1 or higher — ServiceMode. The record applies directly to the name queried and carries a list of SvcParams — the actual connection hints. A name can publish several ServiceMode records at different priorities; a client tries the lowest priority number first, falling back to higher numbers if that connection attempt fails, the same lower-number-wins convention MX priority uses for mail.
AliasModeServiceMode
SvcPriority01 to 65535
SvcParamsIgnored if presentThe connection hints
TargetName .Service unavailable (advisory)Use the owner name
Records per nameOne (RFC says SHOULD)Several, ranked by priority

RFC 9460 §2.4.2 says the main purpose of AliasMode is aliasing at the zone apex, where a CNAME isn’t allowed. Unlike a CNAME, it affects only this one record type, so the apex keeps its SOA, NS, MX and TXT records. If a name publishes both modes, clients must ignore the ServiceMode records (§2.4.1).

Here is the pattern from §10.4.2 and §10.4.3, rewritten with documentation names: an apex aliased to a hosting provider, whose pool offers HTTP/3 and a backup pool on another port.

example.com.       7200 IN HTTPS 0 pool.example.net.
example.com.        300 IN A     192.0.2.1
example.com.        300 IN AAAA  2001:db8::1

pool.example.net.  7200 IN HTTPS 1 . alpn=h2,h3
pool.example.net.  7200 IN HTTPS 2 backup.example.net. alpn=h2 port=8443

A client that understands HTTPS records follows the alias to the pool and tries HTTP/3 there. One that doesn’t uses the A and AAAA records at the apex.

Which SvcParams can a record carry?

Seven keys come from RFC 9460 and its companion ECH document; more have been registered since. The parameters are where the useful information lives. IANA’s Service Binding (SVCB) Parameter Registry, maintained alongside RFC 9460, defines the common keys:

KeyNo.What it says
mandatory0Keys a client must understand to use this record.
alpn1Extra protocols offered, such as h2 or h3.
no-default-alpn2The default protocol is not offered.
port3The port to connect to instead of the default.
ipv4hint4IPv4 addresses for the target.
ech5Encrypted ClientHello configuration (RFC 9848).
ipv6hint6IPv6 addresses for the target.
  • alpn — the ALPN protocol IDs the service supports, such as h2 or h3. This is the headline use case: a client can see a server supports HTTP/3 over QUIC before ever opening a TCP connection to negotiate it the slower way, inside a TLS handshake, as ALPN (RFC 7301) normally requires. As MDN’s own HTTP/3 glossary entry notes, HTTP/3 runs over QUIC rather than TCP at all, which is exactly why advertising support in DNS — before any connection, TCP or otherwise, is opened — is useful.
  • no-default-alpn — a flag telling the client not to assume plain HTTP/1.1 is also offered alongside whatever alpn lists. For HTTPS records the default is http/1.1 (RFC 9460 §9).
  • port — a non-default port the service actually listens on, letting a client skip straight to the right port instead of assuming 443.
  • ipv4hint / ipv6hint — addresses the client may use to start connecting while its own A/AAAA lookup for the target is still pending. RFC 9460 §7.3 tells clients to prefer real A/AAAA answers once they have them, and tells operators not to add hints when the target is the record’s own name.
  • ech — an Encrypted Client Hello configuration, letting the TLS handshake’s SNI field itself be encrypted rather than sent in the clear. The value is a base64 ECHConfigList (RFC 9848 §3); ECH itself is specified in RFC 9849.

None of these are required, and a record can mix and match whichever subset the operator wants to publish. For HTTPS records, port and no-default-alpn are “automatically mandatory” (§9): a client that doesn’t implement them must skip the record rather than connect with the wrong assumptions.

How do you read dig output for an HTTPS record?

Read it left to right: name, TTL, class, type, then the three fields. Example output (illustrative, documentation addresses):

$ dig +noall +answer example.com HTTPS
example.com.        300   IN  HTTPS  1 . alpn="h3,h2"

$ dig +noall +answer www.example.com HTTPS
www.example.com.    300   IN  HTTPS  0 cdn.example.net.

The first answer is ServiceMode (1) with target ., so the endpoint is example.com itself. It offers HTTP/3 and HTTP/2, plus the implied HTTP/1.1, on the usual port. The client still gets the addresses from A/AAAA. The second is AliasMode: the client should repeat the HTTPS query at cdn.example.net and use what it finds there. An ech= parameter would show up as a long base64 string, and a port= parameter as a plain number.

How do clients use an HTTPS record?

They read it before the first connection and adjust how they connect. RFC 9460 describes several effects:

  • Protocol choice. A client compares alpn with what it supports and can open a QUIC connection for HTTP/3 straight away, alongside or instead of TCP (§7.1.2).
  • Upgrade to HTTPS. If an HTTPS record exists, a client about to load an http:// URL should behave as if it had received a 307 redirect to https:// (§9.5).
  • Stricter error handling. For origins with an HTTPS record, a client may remove the option to click through a certificate error (§9.5).
  • ECH. With an ech key in every endpoint, an ECH-capable client must not fall back to a plain connection that would reveal the server name (RFC 9848 §5.1).

The record doesn’t change the origin. A client following an alias to pool.example.net still validates the certificate for example.com (§10.4.2).

How is SVCB used to find encrypted DNS resolvers?

SVCB itself, not HTTPS, describes DNS servers. RFC 9461 maps SVCB to DNS: records live under a _dns prefix, alpn lists dot, doq, h2 or h3, and a dohpath key gives the DoH URI template. The default port is 853 for DoT and DoQ, and 443 for DoH (§4.2).

RFC 9462, Discovery of Designated Resolvers (DDR), uses this so a device can find out whether the plain resolver it was given also offers encrypted DNS. It queries SVCB for the special name _dns.resolver.arpa (§4). Example answer, in the shape of the RFC’s own examples:

_dns.resolver.arpa.  7200 IN SVCB 1 dns.example.net. alpn=h2 dohpath=/dns-query{?dns}
_dns.resolver.arpa.  7200 IN SVCB 2 dns.example.net. alpn=dot

A client should then check that the encrypted resolver’s certificate contains the original resolver’s IP address before switching to it automatically (§4.2, Verified Discovery). For trade-offs between the two transports, see DoH vs. DoT.

Which mistakes break an SVCB or HTTPS record?

Most failures come from records that promise more than the servers deliver.

MistakeWhat happensFix
HTTPS record on a site that only serves plain HTTPClients upgrade http:// to https:// and fail (§9.5)Serve HTTPS first, then publish.
Parameters on an AliasMode recordClients ignore them (§2.4.2)Put hints on the target’s ServiceMode record.
AliasMode and ServiceMode at the same nameServiceMode records are ignored (§2.4.1)Use one mode per name.
alpn=h3 with no QUIC listener or UDP 443 blockedConnections slow down or fail if the client has no TCP fallback (§7.1.2)Advertise h3 only where QUIC answers.
no-default-alpn on every recordClients may discard the whole set (§7.1.2)Keep one record that offers the default protocol.
Stale ipv4hint after moving serversClients can open early connections to the old address (§7.3)Update hints with A/AAAA, or drop them.
A/AAAA removed because HTTPS existsClients without HTTPS-record support can’t connectKeep A/AAAA alongside (§2.4.2).

How do you check for an SVCB or HTTPS record?

Query it with a command-line resolver. AdminKit’s live DNS lookup tool resolves A, AAAA, MX, TXT, NS, CAA, PTR, SOA and CNAME — SVCB and HTTPS are newer types it doesn’t query, and a request for either is rejected before reaching the upstream resolver rather than returning a misleading empty result. To check for one, query it directly with a command-line resolver:

dig example.com HTTPS
dig example.com SVCB

An empty answer section means the domain hasn’t published one, which is completely normal — the record is optional and additive. A domain behind a CDN may have one it never created: Cloudflare’s DNS record types reference says Cloudflare automatically generates HTTPS records for proxied names when Universal SSL is enabled, and serves those instead of manually added HTTPS records on the same names. If you’re confirming a server’s actual TLS behavior rather than what it advertises in DNS, AdminKit’s live SSL/TLS check completes a real handshake against the host and reports the certificate and negotiated protocol version it presents.

Why is publishing one safe for older clients?

Because clients that don’t know the type never act on it. RFC 9460 is explicit that both types are designed to be additive: a resolver or client that has never heard of SVCB or HTTPS simply doesn’t ask for the type, or ignores an answer it happens to receive, and falls back to resolving A/AAAA the way it always has. That is why an operator can add an HTTPS record next to existing A/AAAA records without waiting for every browser, library and resolver to support it; the RFC’s apex example (§10.4.2) notes that clients unaware of the type “will just ignore the new record.” If a change you’ve made doesn’t show up yet in a dig query, see DNS propagation and TTL explained for how long a cached answer can linger before resolvers re-check.

This guide is general information, not a security audit of your specific network — see our disclaimer for what AdminKit's guidance does and does not commit to.

Frequently asked questions

What's the difference between an SVCB record and an HTTPS record?

They share the same wire format and field layout, both defined in RFC 9460. SVCB is the general-purpose record, usable for any service once a matching scheme is defined; HTTPS is a fixed specialization of SVCB for web traffic specifically. A client resolving `https://example.com` looks up the HTTPS record type at that name rather than a generic SVCB record, which is why the two types exist separately instead of one doing both jobs.

What do the priority and target fields actually mean?

RFC 9460 defines two forms based on the SvcPriority field. A priority of 0 is AliasMode: the record carries only a TargetName, telling the client to go resolve that other name's own SVCB/HTTPS record instead — similar in spirit to how a CNAME redirects a lookup, but specific to this record type. Any priority above 0 is ServiceMode: the record applies directly to the name queried, includes a list of SvcParams (the actual hints), and a lower priority number is preferred over a higher one when a name publishes more than one ServiceMode record — the same lower-wins convention MX priority uses.

What hints can an SvcParam actually carry?

RFC 9460 and IANA's Service Binding (SVCB) Parameter Registry define keys including `alpn` (which application protocols the service supports, such as h2 or h3), `no-default-alpn` (don't assume HTTP/1.1 is also offered), `port` (a non-default port the service actually listens on), `ipv4hint`/`ipv6hint` (addresses a client may use to start connecting before its own A/AAAA lookup returns), and `ech` (an Encrypted Client Hello configuration for the TLS handshake). A record can publish any combination of these; none are mandatory.

Does AdminKit's DNS lookup tool show SVCB or HTTPS records?

No. AdminKit's DNS lookup tool resolves A, AAAA, MX, TXT, NS, CAA, PTR, SOA and CNAME — SVCB and HTTPS are newer record types the tool doesn't query, and a request for either is rejected before it reaches the upstream resolver rather than silently returning an empty result. To check for one, query it directly with a command-line resolver, for example `dig example.com HTTPS`.

If a client doesn't understand this record type, what happens?

Nothing breaks. RFC 9460 is explicit that both record types are additive: a resolver or client that doesn't recognize SVCB or HTTPS simply doesn't ask for it, or ignores it if it happens to receive one, and falls back to resolving A/AAAA the same way it always has. This additive design is exactly why large sites have been able to start publishing HTTPS records without waiting for every client on the internet to support them first.

Want AdminKit to watch this for you?

Free checks tell you the state today. Protective DNS and monitoring tell you when it changes.

Huy Nguyen (David), Founder and CEO of Netbase JSC

Talk to an expert

Huy Nguyen (David), Founder & CEO, Netbase JSC

Ask whether this fix applies to your setup, or get a second pair of eyes on your DNS and mail records.