dns_erlang

# Changelog

All notable changes to this project will be documented in this file.

The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## main

## v5.0.18

### Updated

- Name compression is 1.1x-8.9x faster across the benchmark corpus depending on how much a message repeats its names. Both the relative and fully-qualified spelling of a repeated name take the same fast path.
- `t:dns_domain:compmap/0` is now explicitly opaque. Build one with `dns_domain:new_compmap/0`.

## v5.0.17

### Fixed

- `gen_nsec/1,3,4` no longer crashes when a zone carries one name spelled both as an FQDN and relatively (`example.com` and `example.com.`).

## v5.0.16

### Fixed

- `decode_query/1` no longer rejects rfc1995 IXFR queries as `formerr` but as `notimp`.

## v5.0.15

### Fixed

- `decode_message/1` now decodes rfc2136 UPDATE messages, while `decode_query/1` does not anymore as it shouldn't have, returning `notimp`.

## v5.0.14

### Updated

- Optimized `dns_domain:to_wire/3` (compressed name encoding): 1.4–2.8x faster on typical names, up to 17–22x on many-label names such as IPv6 reverse PTRs, with 1.1–3x less allocation.
- Optimized `dns_domain:from_wire/1,2` (wire-format name decoding): ~1.1–1.2x faster on typical compressed names, up to 1.4–1.5x on pointer chains, with up to 18% less allocation on pointer-heavy names.
- Optimized message encoding: ~1.1–1.3x faster on typical bounded responses with 5–35% less allocation across message shapes.

### Fixed

- Correct name compression in the additional section of `encode_message/2` for non-EDNS responses
- NSEC/NSEC3/CSYNC/NXT type bitmap encoding shifted every type after the first by one bit position in any window whose first present type is a multiple of 256
- MINFO records failed to decode inside full messages (`formerr`) due to swapped `dns_domain:from_wire/2` arguments in the MINFO rdata decoder
- `encode_message/2` subtracted the question section twice when sizing the OPT RR and additional records, wrongly dropping them (including the EDNS OPT RR, contra RFC 6891) from responses within question-size bytes of `max_size`
- LOC latitude and longitude used `2^31 - 1` instead of `2^31` as the equator/prime-meridian reference point (RFC 1876 §2)
- LOC `size`/`horiz`/`vert` outside the representable 0–9e9 cm range now raise `badarg`, and in-range values round to nearest instead of truncating (RFC 1876 §2)
- `encode_message/2` with `tc_mode => llq_event` raised `badmatch` when the authority/additional tail or the question section exceeded `max_size`; it now truncates and sets TC
- Zero-length `<character-string>`s were dropped when encoding TXT, SPF, RESINFO and WALLET records (RFC 1035 §3.3)
- A NAPTR record whose REGEXP is not valid UTF-8 decoded into a record holding an error tuple instead of being rejected (RFC 3403 §4.1)
- An EDNS OPT RR advertising a UDP payload size below 512 is clamped to 512 instead of raising `badarg` (RFC 6891 §6.2.3)
- `encode_message/2` with an unsupported `tsig` algorithm raised `case_clause` instead of `badarg`
- An SVCB/HTTPS `alpn` id longer than 255 bytes silently truncated its length octet, producing undecodable wire data; it now raises `{svcb_invalid_alpn_id, _}`
- SVCB/HTTPS `mandatory`, `ipv4hint` and `ipv6hint` values that are not a whole number of elements are rejected instead of silently dropping the remainder (RFC 9460 §7.4, §8)
- An EDNS OPT RR with a non-root owner name, and a message carrying more than one OPT RR, now decode to `formerr` (RFC 6891 §6.1.1)
- `RDLENGTH = 0` for a type this library decodes now decodes to `formerr` instead of an opaque `<<>>` (RFC 1035 §3.2.1)
- An OPT RR anywhere but first in the additional section was ignored when sizing the response and dropped from truncated ones (RFC 6891 §6.1.1)
- URI targets were replaced by their normalized form when decoding, losing data and changing the canonical RDATA that RRSIGs cover (RFC 7553 §4.5)
- A DNSKEY or CDNSKEY with a DSA algorithm but rdata too short to parse as one crashed on re-encode (RFC 2536 §2)
- Re-encoding a DSA DNSKEY whose P began with a zero byte truncated G and Y, producing an undecodable record (RFC 2536 §2)
- `encode_rrdata/2` crashed with `badmap` on every KX record, and with it the DNSSEC canonical form (RFC 2230)
- A repeated or out-of-order NSEC/NSEC3/CSYNC type-bitmap window decoded to a types list containing duplicates (RFC 4034 §4.1.2)
- LOC precision octets with a base nibble above nine were accepted instead of being left opaque (RFC 1876 §2)
- `encode_message/1,2` would emit a message carrying two OPT RRs, which the decoder rejects (RFC 6891 §6.1.1)
- A non-canonically encoded RSA DNSKEY/CDNSKEY public key changed shape when re-encoded, altering the canonical RDATA an RRSIG covers (RFC 3110 §2)

## v5.0.13

### Fixed

- Correctly return `formerr` on invalid NOTIFY DNS messages (rfc1996 §3.7: a NOTIFY carries exactly one question (QDCOUNT=1) and at most one), (#130)

## v5.0.12

### Fixed

- A malformed EDNS COOKIE option (RFC 7873) in a received message now decodes to a `formerr` result instead of raising `bad_cookie`.

## v5.0.11

### Fixed

- WALLET (type 262) RDATA uses TXT-style character-strings on the wire and in zone files (`data` as a list of binaries), per [IANA WALLET](https://www.iana.org/assignments/dns-parameters/WALLET/wallet-completed-template).
- Zone file parsing and RDATA decoding: improved handling of BIND-style presentations and record types;
- Empty TXT/RESINFO RDATA are rejected during semantic validation;
- Empty A/AAAA RDATA is rejected (no longer skipped);

## v5.0.10

### Updated

- Minimal performance improvement for `dns_domain:to_lower/1`, `dns_domain:to_upper/1`, and `dns_domain:are_equal/2`

## v5.0.9

### Updated

- Minimal performance improvement for `dns_domain:are_equal_labels/2`

## v5.0.8

### Fixed

- Handle truncated RR headers in additional and body sections ([#121](https://github.com/dnsimple/dns_erlang/pull/121))

## v5.0.7

### Changed

- Minor internal cleanups in the dnssec module (better function naming)

## v5.0.6

### Added

- `dnssec` exports record types it works over as types.

### Fixed

- Fixed encoding and decoding of ECDSA signatures
- Refactored `dnssec` to use the newer `dns_domain` API internally in a more efficient way
- Incremented test coverage for DNSSEC

### Removed

- Internal ASN1 parser, as it is supported by OTP's `public_key` module.

## v5.0.5

### Fixed

- SVCB/HTTPS params: treat numeric keys key0–key6 as reserved and equivalent to named params (mandatory, alpn, no-default-alpn, port, ipv4hint, ech, ipv6hint). Validate them the same at ingestion in zone parsing, JSON parsing, and wire decoding; key0–key6 are no longer accepted as generic unknown keys ([#110](https://github.com/dnsimple/dns_erlang/issues/110), [#111](https://github.com/dnsimple/dns_erlang/issues/111))
- SVCB params JSON docs: document key0–key6 equivalence to named params, no-default-alpn as null, and keyNNNN format (value binary or null) in generated JSON documentation
- SVCB params wire: decode unknown keys with zero-length value as `none` (not `<<>>`) so round-trip and `to_json` stay consistent (null vs empty binary)
- SVCB params JSON: reject invalid param keys in `to_json/1` (e.g. key > 65535) with `{svcb_param_invalid_key, K}` instead of emitting a map with `undefined` as key
- Ensure strict parsing of IPv4 and IPv6 in zonefiles respectively ([#108](https://github.com/dnsimple/dns_erlang/issues/108))

## v5.0.4

### Fixed

- Fix parsing SVCB json records when `svc_params` are missing

## v5.0.3

### Fixed

- Fix FQDNs an (C)DNSKEY payloads in zonefile encoding

## v5.0.2

### Changed

- Use nulls instead of none in SVCB params JSON payloads [#107](https://github.com/dnsimple/dns_erlang/pull/107)

## v5.0.1

### Fixed

- Fix an issue with the hex package not including all needed files

## v5.0.0

### Added

- Add new `dns_json` module for bidirectional Record <-> JSON/Map transcoding
- Add DNS zone file encoding functionality — convert DNS resource records to RFC 1035 zone file format
- `dns_domain:to_lower/1` and `dns_domain:to_upper/1` for case conversion
- `dns_domain:are_equal/2` and `dns_domain:are_equal_labels/2` for case-insensitive comparison
- `dns_domain:escape_label/1` and `dns_domain:unescape_label/1` for label escaping
- Improved performance with chunked binary pattern matching
- Better RFC1035 and RFC9267 compliance with accurate wire format size tracking

### Changed

- Require OTP 27 or later (`minimum_otp_vsn` set to "27")
- `dns:encode_message/2` changed how truncation is returned to a clearer type
- Migrate domain name operations to new `dns_domain` module

All domain name processing functions have been moved to a new optimized
`dns_domain` module. The old implementations in `dns` and related helper
modules have been removed and replaced with calls to the new module.

#### Migration guide

All domain name functions have been moved to `dns_domain`. Update your code as follows:

```erl
% Old → New
dns:dname_to_lower(Name)             -> dns_domain:to_lower(Name)
dns:dname_to_upper(Name)             -> dns_domain:to_upper(Name)
dns:dname_to_labels(Name)            -> dns_domain:split(Name)
dns:labels_to_dname(Labels)          -> dns_domain:join(Labels)
dns:dname_to_lower_labels(Name)      -> dns_domain:split(dns_domain:to_lower(Name))
dns:compare_dname(NameA, NameB)      -> dns_domain:are_equal(NameA, NameB)
dns:compare_labels(LabelsA, LabelsB) -> dns_domain:are_equal_labels(LabelsA, LabelsB)
dns:escape_label(Label)              -> dns_domain:escape_label(Label)
dns_encode:encode_dname(Name)        -> dns_domain:to_wire(Name)
dns_encode:encode_dname(CM, Pos, N)  -> dns_domain:to_wire(CM, Pos, N)
```

**Note:** The wrapper functions in `dns` module (`dname_to_labels/1`, `labels_to_dname/1`,
`dname_to_lower_labels/1`, `dname_to_upper/1`, `dname_to_lower/1`, `compare_dname/2`,
`compare_labels/2`, `escape_label/1`) have been removed. Use `dns_domain` functions directly now.

**Error Handling Changes:**

- decoding invalid wire packets previously used `throw/1` for errors (`decode_loop`, `bad_pointer`),
  it uses `error/1` now
- Update error handling: `try dns:decode_message(...) catch error:Reason -> ... end`

### Removed

- Removed old `dns_record` and `dns_record_info` modules

## 4.9.1

### Fixed

- Fix a bug that would incorrectly calculate pointers, introduced in 4.9.0

## 4.9.0

### Added

- Add extended support for SVCB and HTTPS Resource Records — [RFC9460](https://datatracker.ietf.org/doc/rfc9460/)
- Add support for OPENPGPKEY (Type 61) — [RFC 7929](https://datatracker.ietf.org/doc/rfc7929/)
- Add support for SMIMEA (Type 53) — [RFC 8162](https://datatracker.ietf.org/doc/rfc8162/)
- Add support for URI (Type 256) — [RFC 7553](https://datatracker.ietf.org/doc/rfc7553/)
- Add support for WALLET (Type 262) — IANA Registration
- Add support for EUI48 (Type 108) and EUI64 (Type 109) — [RFC 7043](https://datatracker.ietf.org/doc/rfc7043/)
- Add support for CSYNC (Type 62) — [RFC 7477](https://datatracker.ietf.org/doc/rfc7477/)
- Add support for DSYNC (Type 66) — [RFC 9859](https://datatracker.ietf.org/doc/rfc9859/)
- Add `dns:decode_query/1` function for strict query validation to prevent DoS attacks

### Changed

- Performance improvements in message encoding (~5-15% faster for encode/1, ~0-20% faster for encode/2)

## 4.8.1

### Fixed

- Fix bad size for reserved OptRR record during encoding

## 4.8.0

### Added

- Add support for [RFC8976](https://datatracker.ietf.org/doc/rfc8914/): Message Digest for DNS Zones

### Fixed

- Fix Base64 lexing of zone files

## 4.7.0

### Added

- Extend support for [RFC8914](https://datatracker.ietf.org/doc/rfc8914/): Extended DNS Errors.

## 4.6.0

### Added

- Add support for DNS zone file parsing

## 4.5.0

### Added

- Add support for [RFC8080](https://datatracker.ietf.org/doc/rfc8080/): Ed22519/Ed448 for DNSSEC.
- Add support for [RFC8914](https://datatracker.ietf.org/doc/rfc8914/): Extended DNS Errors.

## 4.4.0

### Added

- Add support for [RFC6605](https://datatracker.ietf.org/doc/rfc6605/): ECDSA for DNSSEC.
- Add support for [RFC9077](https://datatracker.ietf.org/doc/rfc9077/): NSEC/NSEC3 TTLs.

## 4.3.0

### Added

- Add support for TLSA records.

## 4.2.0

### Added

- Add shortcuts for normalizing and converting domains into labels.
- Add a helper for comparing label lists.
- Add helpers to convert from and to codes into their string representation.

## 4.1.0

### Added

- Add support for RFC7873 EDNS cookies encoding and decoding

## 4.0.1

### Fixed

- Fix additional count when truncating a message

## 4.0.0

### Changed

- Use `erlang:system_time/1` for timestamps.
- Use maps instead of proplists for options to be passed to
    encoding, decoding, TSIG, and DNSSEC functions.
- Encoding logic builds responses that contain the full question section,
    and optionally drop the answers if the response is to be truncated.
- Use consistent record names.
- Split names functions into a standalone `dns_names` module.
- Extends and reorganise documentation.

## 3.1.3

### Fixed

- Fix: now using the reserved space for the OPT RR records during the encoding of the message (#74)

## 3.1.2

### Added

- Add NXNAME type to terms types.

## 3.1.1

### Fixed

- Fix EDNS0 compliance for truncated records and unsupported versions

## 3.1.0

### Fixed

- Fix EDNS0 compliance for truncated records and unsupported versions

## 3.0.5

### Fixed

- Fix `max_size` in encode_message opts

## 3.0.4

### Changed

- Upgrade dependencies
- Apply linter specs

## 3.0.3

### Changed

- Performance improvements in string manipulations

## 3.0.2

### Fixed

- Fix type definitions

## 3.0.1

### Fixed

- Ensure ASN1 compilation before building package

## 3.0.0

### Added

- Add xref, dialyzer, and ex_doc
- Add strict typing and RFC references to all records
- Add support for TXT splitting of strings over the maximum permitted size

## 2.0.0

### Changed

- Bump to OTP/27
- Replace "jsx" with "json"

### Added

- Erlfmt
- CONTRIBUTING.md
- CHANGELOG.md
- Release process to hex.pm

## 1.1.0

N/A