Hi Paul,

Understood. The resource scarcity argument makes sense. Taking a dedicated 
RRTYPE for `UTXO` is wasteful if many apps just need a binary container.

Revised plan:

1. Drop the `UTXO` RRTYPE request entirely. 
2. Submit a new request for a generic `CONTAINER` RRTYPE. 
3. `_utxo` becomes just one application-specific prefix under this container, 
exactly as you suggested (like `_sip._tcp`).

This is the same pattern ENS uses in DNS (reusing a generic container with an 
application prefix like `ENS1`). One generic type, application semantics 
isolated by prefix.

I'll revise the IANA template under #1495165 to request `CONTAINER` instead, 
and update the draft to reflect the prefix approach.

Let me know if you have any specific naming preferences or if there are any 
existing containers we should reuse before I submit the revised template.

-G



来自阿里邮箱Android客户端------------------------------------------------------------------
发件人:Paul Wouters<[email protected]>
日 期:2026年10月28日 01:31:21
收件人:Foundation<[email protected]>
抄 送:dns-rrtype-applications<[email protected]>; <[email protected]>
主 题:Re: [DNSOP] RRTYPE Application: UTXO (Data RR)

On Tue, 27 Oct 2026, Foundation wrote:

> D. Motivation for the new RRTYPE application.
> 
> Traditional DNS answers only the "where" question—mapping a domain name to an 
> IP address. Modern payment scenarios require the same domain name to 
> simultaneously
> answer three additional questions: who is the receiving party (identity), 
> what rules apply (compliance), and to which endpoint the value should 
> actually be sent
> (payment endpoint). These are DNS's structural blind spots, and no existing 
> RRTYPE is designed to carry this combination of verifiable cryptographic 
> pointers and
> regulatory metadata.

This just means the information should be somewhere in DNS, it does not
really mean it has to be its own RRtype.

> This application requests a Data RRTYPE that functions as a neutral, opaque 
> container for DNS. The RDATA is a CBOR-encoded payload containing a 
> blockchain UTXO
> reference, an optional Verifiable Random Function (VRF) proof, a payment 
> endpoint identifier, and optional vLEI legal-entity references. The DNS core 
> protocol does
> not parse or interpret this payload; wallets and TEE clients decode and 
> verify it locally. The goal is to extend DNS from a one-dimensional 
> name-to-IP mapping into
> a multi-dimensional resolver for identity, compliance, and payment endpoints, 
> while leaving DNS resolution semantics entirely unchanged.

This possibly jsutifies an RRtype that can be used to ship CBOR endcoded
payloads, but then it should not be called UTXO but be kept rather
generic like perhaps CBORDATA, so other applications with similar
requirements could use it without confusion. Usage should be
disambiguated via an application specific _prefix, eg "_utxo".


> E. Description of the proposed RR type.
> 
> The RDATA is an opaque byte stream, CBOR-encoded per RFC 8949. It carries a 
> structured payload that includes a UTXO reference, an endpoint type, a chain 
> identifier,
> an endpoint identifier, an expiry timestamp, an owner public-key hash, and 
> optional fields for a VRF proof (per RFC 9381), an IPv6 Interface Identifier 
> (per RFC
> 7217), and a vLEI DID reference. The complete binary format, field ordering, 
> and client-side parsing rules are specified in draft-guorong-utxo-dns-02, 
> Section 2.3.

The rdata seems to be a single blob, yes. The blob could be UTXO, or
something else.

> F. What existing RRTYPE or RRTYPEs come closest to filling that need and why 
> are they unsatisfactory?
> 
> TXT (type 16) is the closest existing Data RR for carrying arbitrary data. 
> However, TXT is designed for unstructured UTF-8 text; it provides no binary 
> container
> semantics, no type discrimination, and no mechanism to signal that the 
> payload must be cryptographically verified by the client. Encoding a CBOR 
> structure inside
> TXT would require an out-of-band convention that is invisible to DNS 
> infrastructure and indistinguishable from ordinary text records.

I'm not sure I undestand this argument fully, because any binary blob
can be base encoded into a TXT record RRdata portion. Its just another
container layer.

> No existing RRTYPE combines three properties simultaneously: (1) an opaque, 
> client-parsed binary container, (2) explicit signalling that the payload is 
> intended for
> cryptographic verification outside DNS, and (3) coexistence with RFC 3597 
> Unknown RR handling without resolver modifications. A new Data RRTYPE is 
> required to fill
> this gap.

I don't think the signaling of "cryptographic verification outside DNS"
needs to be part of the RRtype definition, and it isn't since as you
say, it is just a blob.

> G. What mnemonic is requested for the new RRTYPE (optional)?
> 
> UTXO

The reason I believe this registration should be generalized is that I
don't think we have enough RRtype entries to make the RRtype mnemonic infer
a bunch of conditions on the blob payload that differ per application. I
am not against a generic CONTAINER RRtype that is used outside the apex
using a per-application globally scoped underscore prefix (eg _utxo)
that can be used by many applications that want to pull some binary out
from DNS, whether it is CBOR encoded or otherwise.

Paul
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to