<?xml version="1.0" encoding="UTF-8"?>
<rfc xmlns:xi="http://www.w3.org/2001/XInclude"
     category="info" docName="draft-toco-abiz-dns-00" ipr="trust200902"
     submissionType="IETF" xml:lang="en" version="3">
  <front>
    <title abbrev="ABIZ DNS Pointer">The "_abiz" Underscored and Globally Scoped DNS Node Name for Agentic Business Records</title>
    <seriesInfo name="Internet-Draft" value="draft-toco-abiz-dns-00"/>
    <author fullname="Peter Chapman" initials="P." surname="Chapman">
      <organization>To.Co Technologies Limited</organization>
      <address><postal><street>66 Paul Street</street><city>London</city><code>EC2A 4NA</code><country>United Kingdom</country></postal><email>pjc@to.co</email></address>
    </author>
    <date year="2026" month="November" day="16"/>
    <area>art</area>
    <keyword>DNS</keyword><keyword>ABIZ</keyword><keyword>business record</keyword><keyword>agent discovery</keyword>
    <abstract><t>This document specifies use of the underscored DNS node name "_abiz" with TXT resource records to publish a pointer from a domain to a signed public business record served over HTTPS. It describes discovery and verification of that record and requests registration of the "_abiz" node name and the "abiz.json" well-known URI suffix.</t></abstract>
  </front>
  <middle>
    <section anchor="introduction" numbered="true">
      <name>Introduction</name>
      <t>Businesses publish public facts about themselves on the web, but software that discovers a domain may not have a provider-independent way to locate a structured, signed record of those facts. ABIZ (Agentic Business Record) uses a TXT record at "_abiz" under a domain to point to an HTTPS resource serving a signed JSON business record.</t>
      <t>This document describes the DNS assertion, the discovery procedure, the checks that link a record to the queried domain, and the registrations. The ABIZ 1.1 record schema and the rules for the record's content are specified in <xref target="ABIZSPEC"/>. An ABIZ assertion does not itself establish legal identity, business registration, or endorsement by an AI provider.</t>
    </section>
    <section anchor="terminology">
      <name>Terminology</name>
      <t>The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 <xref target="RFC2119"/> <xref target="RFC8174"/> when, and only when, they appear in all capitals, as shown here.</t>
      <dl>
        <dt>ABIZ assertion:</dt><dd>A versioned tag list carried in a TXT resource record at "_abiz.&lt;domain&gt;", or the same values carried in a well-known document (<xref target="well-known"/>).</dd>
        <dt>Queried domain:</dt><dd>The domain name for which discovery is being performed, written "&lt;domain&gt;".</dd>
        <dt>Record host:</dt><dd>The host, optionally followed by a permitted path, identified by the "h" tag.</dd>
        <dt>Producer:</dt><dd>The party that publishes an ABIZ assertion for a domain it controls.</dd>
        <dt>Consumer:</dt><dd>Software that discovers and validates an ABIZ record.</dd>
      </dl>
    </section>
    <section anchor="dns-assertion"><name>The ABIZ DNS Assertion</name>
      <section anchor="owner-name"><name>Owner Name and RRset</name>
        <t>The producer publishes the assertion in a TXT resource record at "_abiz.&lt;domain&gt;". A TXT record is an ABIZ assertion when its first tag is "v=abiz1" (<xref target="tags"/>). There <bcp14>MUST</bcp14> be exactly one such record at that owner name. More than one makes the ABIZ assertion invalid; none means that no DNS assertion is present. Other TXT records at the same name (including one whose first tag is "V=ABIZ1" or "v=abiz1x") are not ABIZ assertions and are not counted.</t>
        <t>The character-strings within a single TXT resource record <xref target="RFC1035"/> are concatenated in order, with nothing added between them, before tag parsing. The TTL <bcp14>SHOULD</bcp14> be 300 seconds.</t>
      </section>
      <section anchor="tags"><name>Tag Syntax</name>
        <t>After character-string concatenation, the assertion <bcp14>MUST</bcp14> match "abiz-record" in the following ABNF <xref target="RFC5234"/> <xref target="RFC7405"/>:</t>
        <sourcecode type="abnf"><![CDATA[abiz-record  = *WSP abiz-version *( abiz-sep abiz-tag )
               [ *WSP ";" ] *WSP
abiz-version = %s"v" *WSP "=" *WSP %s"abiz1"
abiz-sep     = *WSP ";" *WSP
abiz-tag     = tag-name *WSP "=" *WSP tag-value
tag-name     = 1*( %x21-3A / %x3C / %x3E-7E / %x80-FF )
tag-value    = 1*( %x21-3A / %x3C-7E / %x80-FF )
WSP          = SP / HTAB]]></sourcecode>
        <ul>
          <li>Space and horizontal tab are the only whitespace, and are permitted only at the start and end of the assertion and around ";" and "=". Other control characters <bcp14>MUST NOT</bcp14> appear. A tag value does not contain whitespace or ";".</li>
          <li>A tag is split on its first "=" only; a value may itself contain "=".</li>
          <li>At most one trailing ";" is permitted. An empty tag, a tag without "=", and a tag with an empty name or value make the assertion invalid.</li>
          <li>Tag names and the value "abiz1" are case-sensitive. "H=" is an unknown tag, not "h=".</li>
          <li>A recognised tag ("v", "h", "k", "p") that occurs more than once makes the assertion invalid. Unknown tags <bcp14>MUST</bcp14> be ignored, repeated or not, and <bcp14>MUST NOT</bcp14> be given any meaning.</li>
        </ul>
        <t>Producers <bcp14>SHOULD</bcp14> publish the compact form with no optional whitespace. Consumers <bcp14>MUST</bcp14> accept the full syntax.</t>
      </section>
      <section anchor="tag-defs"><name>Tags</name>
        <dl>
          <dt>v (<bcp14>REQUIRED</bcp14>):</dt><dd>Protocol family version, with the exact value "abiz1". It is always the first tag.</dd>
          <dt>h (<bcp14>REQUIRED</bcp14>):</dt><dd>The record host: a DNS host name, without scheme, user information, port, query, or fragment. A path <bcp14>MAY</bcp14> follow the host name, only to identify a business unit or division as permitted by <xref target="ABIZSPEC"/>; it <bcp14>MUST NOT</bcp14> contain "." or ".." segments. The host name <bcp14>MUST</bcp14> be a public name. Consumers <bcp14>MUST</bcp14> reject an "h" that names a special-use, loopback, private-network, or link-local host. The record host <bcp14>MAY</bcp14> be operated by a party other than the producer.</dd>
          <dt>p (<bcp14>REQUIRED</bcp14>):</dt><dd>The raw 32-byte Ed25519 public key <xref target="RFC8032"/> in the base64 encoding of <xref target="RFC4648" section="4"/>, with padding: exactly 44 characters, the last of which is "=". The encoding <bcp14>MUST</bcp14> be canonical (encoding the decoded octets again gives the same text). The base64url alphabet, unpadded values, and hexadecimal strings are invalid.</dd>
          <dt>k (<bcp14>OPTIONAL</bcp14>):</dt><dd>The business registration key, in the form specified by <xref target="ABIZSPEC"/>. It is compared case-insensitively.</dd>
        </dl>
        <t>An assertion that lacks "h" or "p", or in which "h", "p", or a present "k" is malformed, is invalid.</t>
      </section>
      <section anchor="examples"><name>Examples</name>
        <t>Registered business (illustrative data):</t>
        <sourcecode type="dns-rr"><![CDATA[_abiz.harbourbikes.example. 300 IN TXT (
  "v=abiz1;k=us-or-123456;"
  "h=harbour-bike-repair.or.us.to.co;"
  "p=oQ0t/PFgN9pH0QSNUHrY5FLj5igISyzD1z6AXr5NtZo=")]]></sourcecode>
        <t>Unregistered business (illustrative data):</t>
        <sourcecode type="dns-rr"><![CDATA[_abiz.marketstand.example. 300 IN TXT (
  "v=abiz1;h=market-stand-repairs-portland.biz.to.co;"
  "p=zc6S6bKXaZXO7NCI+fjtgMWLdbdEeavjVvydQsoQCJ0=")]]></sourcecode>
        <t>The examples are fictional; the public keys are demonstration values and <bcp14>MUST NOT</bcp14> be reused in production.</t>
      </section>
    </section>
    <section anchor="well-known"><name>Well-Known URI Alternative</name>
      <t>A producer that cannot publish a TXT record, or prefers not to, <bcp14>MAY</bcp14> publish the same assertion at "https://&lt;domain&gt;/.well-known/abiz.json" <xref target="RFC8615"/>, where "&lt;domain&gt;" is exactly the queried domain (no "www." added or removed) on the default HTTPS port. A producer <bcp14>MAY</bcp14> publish both; if it does, the two <bcp14>MUST</bcp14> carry identical values.</t>
      <t>The document is a UTF-8 JSON object <xref target="RFC8259"/> whose members "v", "h", "p", and the optional "k" are JSON strings holding exactly the corresponding tag values. A member that is not a string, or whose value would not be a valid tag value, makes the document invalid. Unknown members <bcp14>MUST</bcp14> be ignored. The document is a pointer, not the business record.</t>
      <sourcecode type="json"><![CDATA[{
  "v": "abiz1",
  "k": "us-or-123456",
  "h": "harbour-bike-repair.or.us.to.co",
  "p": "oQ0t/PFgN9pH0QSNUHrY5FLj5igISyzD1z6AXr5NtZo="
}]]></sourcecode>
      <t>The producer <bcp14>MUST</bcp14> serve the document with status 200, without a redirect, with the media type "application/json", and at no more than 65,536 octets. A consumer <bcp14>MUST</bcp14> apply the safe-retrieval rules of <xref target="retrieval"/>, <bcp14>MUST NOT</bcp14> follow redirects, and <bcp14>MUST</bcp14> treat any other status, media type, or a larger body as an invalid document. A consumer <bcp14>MUST NOT</bcp14> cache the document for more than 300 seconds.</t>
    </section>
    <section anchor="discovery"><name>Discovery Procedure</name>
      <t>The DNS assertion takes precedence over the well-known document. A consumer:</t>
      <ol>
        <li><bcp14>MUST</bcp14> first query TXT at "_abiz.&lt;domain&gt;".</li>
        <li>If the query fails (for example, SERVFAIL or a timeout), <bcp14>MUST</bcp14> report the result as unavailable. It <bcp14>MUST NOT</bcp14> report that no assertion exists and <bcp14>MUST NOT</bcp14> use the well-known document in its place.</li>
        <li>If the RRset holds two or more ABIZ assertions, <bcp14>MUST</bcp14> treat the result as invalid and <bcp14>MUST NOT</bcp14> use the well-known document.</li>
        <li>If it holds exactly one, <bcp14>MUST</bcp14> use that assertion. If the assertion is malformed, the result is invalid and the well-known document <bcp14>MUST NOT</bcp14> be used in its place. A consumer <bcp14>MAY</bcp14> also fetch the well-known document; if that document disagrees with the DNS assertion or is invalid, the consumer <bcp14>MUST</bcp14> use the DNS assertion and <bcp14>SHOULD</bcp14> report a warning.</li>
        <li>If there is no ABIZ assertion in DNS (a name error, no data, or only other TXT records), <bcp14>SHOULD</bcp14> fetch the well-known document. If it is valid, its members are treated exactly as the tags of a DNS assertion. If it is absent or invalid, there is no ABIZ assertion for the domain.</li>
      </ol>
      <t>A consumer <bcp14>MUST</bcp14> record which source it used. Only the queried domain is consulted: a consumer <bcp14>MUST NOT</bcp14> substitute a parent domain or another name.</t>
    </section>
    <section anchor="record"><name>Relationship to the Public Business Record</name>
      <section anchor="retrieval"><name>Retrieval</name>
        <t>The consumer retrieves the record with a GET request to "https://&lt;h&gt;", sending "Accept: application/json"; it <bcp14>SHOULD</bcp14> also append the query "format=json" as defined in <xref target="ABIZSPEC"/>. When retrieving the record or the well-known document, a consumer <bcp14>MUST</bcp14>:</t>
        <ul>
          <li>use HTTPS with certificate and host name validation;</li>
          <li>refuse to connect to loopback, private-network, and link-local addresses, checking the addresses the host name resolves to;</li>
          <li>not follow unsafe redirects (a consumer <bcp14>MAY</bcp14> refuse all redirects, and <bcp14>MUST</bcp14> refuse them for the well-known document);</li>
          <li>limit the response size and the time it waits;</li>
          <li>accept only a 200 response whose media type is "application/json", and parse it as data without executing anything.</li>
        </ul>
        <t>A failure that may be temporary (a timeout, a connection or TLS failure, or a 408, 429, or 5xx status) does not establish that the record is absent or invalid; a consumer <bcp14>SHOULD</bcp14> report it as unavailable.</t>
      </section>
      <section anchor="verification"><name>Verification</name>
        <t>Before relying on a retrieved record, a consumer <bcp14>MUST</bcp14> check all of the following. A record that fails any of them <bcp14>MUST NOT</bcp14> be treated as published by the controller of the queried domain.</t>
        <ol>
          <li>The body is a JSON object in the I-JSON profile <xref target="RFC7493"/>; a repeated member name is an error. It validates against the ABIZ 1.1 schema <xref target="ABIZSPEC"/>.</li>
          <li>The record's "identity_host" member equals the "h" value exactly, including any path.</li>
          <li>The record's "domains" member includes the queried domain.</li>
          <li>If the assertion has "k", the record has a "business_key" member equal to it, compared case-insensitively.</li>
          <li>The record's "public_key" member equals the "p" value (an exact string comparison, since both are canonical).</li>
          <li>The record's "signature" member is a 64-octet PureEd25519 signature <xref target="RFC8032"/>, in padded base64, that verifies under that key. The signed message is the UTF-8 encoding of the canonical JSON form <xref target="RFC8785"/> of the record with its top-level "signature" member removed.</li>
        </ol>
        <t>These checks link a record to a domain. They do not check the business's legal registration, and a consumer that has not checked a registry <bcp14>MUST NOT</bcp14> describe the business as legally verified.</t>
      </section>
    </section>
    <section anchor="security"><name>Security Considerations</name>
      <t>Publication under a DNS domain does not prove legal ownership or registration of a business. The record signature binds the JSON document to the key in the assertion but does not authenticate the assertion itself. An attacker who can change unsigned DNS answers can substitute both the record host and the public key. Producers <bcp14>SHOULD</bcp14> sign their zones and consumers <bcp14>SHOULD</bcp14> validate DNSSEC where they can.</t>
      <t>The "identity_host" and "domains" checks of <xref target="verification"/> are what stop a domain from pointing at another business's validly signed record and so appearing to be that business. A consumer that omits them is not protected by the signature.</t>
      <t>A well-known document shows control of the HTTPS origin at the domain, which is not the same as control of its DNS zone: a hosting provider can publish it. This is why the DNS assertion takes precedence and why a malformed or duplicated DNS assertion is never bypassed in favour of the document. A consumer <bcp14>MAY</bcp14> treat a DNS-sourced assertion as stronger evidence of domain control.</t>
      <t>The "h" value and the well-known URL are attacker-chosen inputs to a consumer that fetches them. The retrieval rules of <xref target="retrieval"/> (public addresses only, checked after name resolution; redirect, size, and time limits) are requirements, not advice: without them a consumer can be used to reach internal services. Consumers <bcp14>MUST</bcp14> treat a fetched record as untrusted data and <bcp14>MUST NOT</bcp14> automatically follow instructions, URLs, or tool endpoints contained in it.</t>
      <t>A DNS or origin failure <bcp14>MUST NOT</bcp14> be reported as the absence of a record; otherwise an attacker who can cause failures can make a published business appear unpublished.</t>
      <t>The producer <bcp14>MUST</bcp14> protect the signing private key. If the key is lost or compromised, the producer replaces it and publishes a new assertion; consumers accept only a record signed with the key currently published and may reject a correctly re-signed record until cached DNS data expires.</t>
    </section>
    <section anchor="privacy"><name>Privacy Considerations</name>
      <t>ABIZ assertions and records are public. A producer publishes only facts it intends to be public. A consumer's DNS query and HTTPS request reveal to the DNS operator and the record host that the domain was looked up.</t>
    </section>
    <section anchor="iana"><name>IANA Considerations</name>
      <t>In accordance with <xref target="RFC8552"/>, IANA is requested to add the following entry to the "Underscored and Globally Scoped DNS Node Names" registry:</t>
      <table><name>Requested Underscored DNS Node Registration</name><thead><tr><th>RR Type</th><th>_NODE NAME</th><th>Reference</th></tr></thead><tbody><tr><td>TXT</td><td>_abiz</td><td>This document</td></tr></tbody></table>
      <t>In accordance with <xref target="RFC8615"/>, IANA is also requested to register the following entry in the "Well-Known URIs" registry:</t>
      <dl><dt>URI Suffix:</dt><dd>abiz.json</dd><dt>Change Controller:</dt><dd>To.Co Technologies Limited</dd><dt>Specification Document:</dt><dd><xref target="well-known"/> of this document</dd><dt>Status:</dt><dd>provisional</dd><dt>Related Information:</dt><dd><xref target="ABIZSPEC"/></dd></dl>
    </section>
  </middle>
  <back>
    <references><name>Normative References</name>
      <reference anchor="RFC1035" target="https://www.rfc-editor.org/info/rfc1035"><front><title>Domain names - implementation and specification</title><author initials="P." surname="Mockapetris" fullname="Paul Mockapetris"/><date year="1987" month="November"/></front><seriesInfo name="STD" value="13"/><seriesInfo name="RFC" value="1035"/></reference>
      <reference anchor="RFC2119" target="https://www.rfc-editor.org/info/rfc2119"><front><title>Key words for use in RFCs to Indicate Requirement Levels</title><author initials="S." surname="Bradner" fullname="Scott Bradner"/><date year="1997" month="March"/></front><seriesInfo name="BCP" value="14"/><seriesInfo name="RFC" value="2119"/></reference>
      <reference anchor="RFC4648" target="https://www.rfc-editor.org/info/rfc4648"><front><title>The Base16, Base32, and Base64 Data Encodings</title><author initials="S." surname="Josefsson" fullname="Simon Josefsson"/><date year="2006" month="October"/></front><seriesInfo name="RFC" value="4648"/></reference>
      <reference anchor="RFC5234" target="https://www.rfc-editor.org/info/rfc5234"><front><title>Augmented BNF for Syntax Specifications: ABNF</title><author initials="D." surname="Crocker" fullname="Dave Crocker" role="editor"/><author initials="P." surname="Overell" fullname="Paul Overell"/><date year="2008" month="January"/></front><seriesInfo name="STD" value="68"/><seriesInfo name="RFC" value="5234"/></reference>
      <reference anchor="RFC7405" target="https://www.rfc-editor.org/info/rfc7405"><front><title>Case-Sensitive String Support in ABNF</title><author initials="P." surname="Kyzivat" fullname="Paul Kyzivat"/><date year="2014" month="December"/></front><seriesInfo name="RFC" value="7405"/></reference>
      <reference anchor="RFC7493" target="https://www.rfc-editor.org/info/rfc7493"><front><title>The I-JSON Message Format</title><author initials="T." surname="Bray" fullname="Tim Bray" role="editor"/><date year="2015" month="March"/></front><seriesInfo name="RFC" value="7493"/></reference>
      <reference anchor="RFC8032" target="https://www.rfc-editor.org/info/rfc8032"><front><title>Edwards-Curve Digital Signature Algorithm (EdDSA)</title><author initials="S." surname="Josefsson" fullname="Simon Josefsson"/><author initials="I." surname="Liusvaara" fullname="Ilari Liusvaara"/><date year="2017" month="January"/></front><seriesInfo name="RFC" value="8032"/></reference>
      <reference anchor="RFC8174" target="https://www.rfc-editor.org/info/rfc8174"><front><title>Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words</title><author initials="B." surname="Leiba" fullname="Barry Leiba"/><date year="2017" month="May"/></front><seriesInfo name="BCP" value="14"/><seriesInfo name="RFC" value="8174"/></reference>
      <reference anchor="RFC8259" target="https://www.rfc-editor.org/info/rfc8259"><front><title>The JavaScript Object Notation (JSON) Data Interchange Format</title><author initials="T." surname="Bray" fullname="Tim Bray" role="editor"/><date year="2017" month="December"/></front><seriesInfo name="STD" value="90"/><seriesInfo name="RFC" value="8259"/></reference>
      <reference anchor="RFC8552" target="https://www.rfc-editor.org/info/rfc8552"><front><title>Scoped Interpretation of DNS Resource Records through "Underscored" Naming of Attribute Leaves</title><author initials="D." surname="Crocker" fullname="Dave Crocker"/><date year="2019" month="March"/></front><seriesInfo name="BCP" value="222"/><seriesInfo name="RFC" value="8552"/></reference>
      <reference anchor="RFC8615" target="https://www.rfc-editor.org/info/rfc8615"><front><title>Well-Known Uniform Resource Identifiers (URIs)</title><author initials="M." surname="Nottingham" fullname="Mark Nottingham"/><date year="2019" month="May"/></front><seriesInfo name="RFC" value="8615"/></reference>
      <reference anchor="RFC8785" target="https://www.rfc-editor.org/info/rfc8785"><front><title>JSON Canonicalization Scheme (JCS)</title><author initials="A." surname="Rundgren" fullname="Anders Rundgren"/><author initials="B." surname="Jordan" fullname="Bret Jordan"/><author initials="S." surname="Erdtman" fullname="Samuel Erdtman"/><date year="2020" month="June"/></front><seriesInfo name="RFC" value="8785"/></reference>
      <reference anchor="ABIZSPEC" target="https://github.com/abiz-org/abiz-agentic-business-record/blob/v1.1.0/SPEC.md"><front><title>ABIZ - Agentic Business Record Specification, Version 1.1</title><author><organization>To.Co Technologies Limited</organization></author><date year="2026" month="November"/></front></reference>
    </references>
  </back>
</rfc>
