v1.4.0

Fixes from running the library over public SDP sets; see "Checked against public SDP sets" in
the README.

Added

• yarn check:external runs the library over public collections and captures, downloaded into
  .cache/. It is a manual check, not part of the tests or CI, and none of the data is in this
  repository.

Fixed

• sdpReader.parse accepted an m= line with an empty media type, proto or non-RTP format
  (m=audio 9 RTP/AVP 0 gave proto: "" and lost the codecs). It now throws
  expected "<media> <port> <proto> <fmt> ...". Strict mode rejects input it used to accept.
• sdpReader.parseLenient: doubled spaces in a malformed m= line shifted the fields
  (m=audio 9 RTP/AVP 0 gave proto: "9" and no port) or left '' in formats.
• sdpReader.isComplete is false for a media section without formats (m=video 0 RTP/AVP),
  which sdpWriter.write refuses.

v1.3.0

Added

• Lenient parsing: sdpReader.parseLenient(text) never throws, keeps everything the text
  allows and returns {sdp, warnings}. Each deviation that parse rejects becomes a
  warning {code, reason, lineNumber, line, action} with the same reason (and, for the
  first warning, the same line) as the strict error. Trailing spaces/tabs (except on
  s=/i=/u=/e=/p=) and CR-only line endings are handled as well. New types
  SdpLenient, SdpLenientMediaSection, SdpLenientResult, SdpParseWarning,
  SdpParseWarningCode, SdpParseWarningAction, and the type guard
  sdpReader.isComplete(sdp). See "Lenient parsing" in the README.
• sdpValidator.validate(text, {lenient: true}) reports parse warnings as
  lenient:<code> issues and checks the recovered model. SdpIssueCode now includes the
  lenient:* codes (a wider union type).
• wasm: sdp_parse_lenient and sdp_validate_lenient (C ABI), parseLenient and
  validateLenient (ESM bindings); @irbisadm/simple-sdp/wasm has parseLenient,
  isComplete and the lenient validator option. Swift: parseLenient(_:),
  parseLenientJSON(_:), validateLenient(_:type:). Kotlin: parseLenient(sdp),
  validateLenient(sdp, type).

Changed

• sdpWriter.write accepts SdpLenient and throws for a media section without port or
  proto (it used to write m=audio undefined …).
• wasm: a model whose media section has no port or proto is accepted by sdp_write
  and sdp_validate_model like in TypeScript (writing fails with the message above);
  it used to fail with "invalid SDP model".

Fixed

• wasm: a description whose last line ends with a lone CR was parsed differently from
  TypeScript (the CR was dropped); the strict parser now matches TypeScript.
• A malformed a=rtpmap no longer leaves a partially filled codec behind (not visible
  through parse, which throws, but used by lenient parsing).

v1.2.1-rc.1

Changed

• Releases are published to npm by CI (release.yml) with npm trusted publishing, so
  every version from now on carries an npm provenance attestation.

v1.2.0

Added

• @irbisadm/simple-sdp/wasm: the same API and types as the main entry point, run by the
  WebAssembly module (ESM with top-level await; works in Node.js and with bundlers that
  handle new URL(…, import.meta.url), checked with Vite 6). The module itself is
  exported as @irbisadm/simple-sdp/wasm/simple-sdp.wasm.
• Kotlin library on JitPack: com.github.irbisadm:simple-sdp:v1.2.0 (jitpack.yml).
• GitHub Releases carry simple-sdp.wasm for other runtimes (.github/workflows/release.yml).

v1.0.1

Fixed

• Validator accepted malformed a=fingerprint values such as sha-256 AA:BB.
  New issue invalid-fingerprint (error) checks that a=fingerprint is colon-separated
  hex pairs with the digest length of its hash function (md2/md5 16, sha-1 20, sha-224 28,
  sha-256 32, sha-384 48, sha-512 64 bytes); fingerprint-not-uppercase (warning).
  RFC 8122 §5.
• package.json: repository in the normalized object form (no npm publish warning).

Install: npm i @irbisadm/simple-sdp@1.0.1

v1.0.0

The model of 0.2.0 is now the stable 1.x API. No breaking changes since 0.2.0.

Added

• sdpWriter.write(sdp): string. Lines are written in RFC 8866 §5 order with CRLF;
  attribute order inside a section is normalized (see README, "Writing"). Round-trip
  tests cover every fixture.
• sdpValidator.validate(sdp | text, {type?}): SdpIssue[], plus types SdpIssue,
  SdpIssueCode and SdpValidateOptions (rules: README, "Validation").
• package.json: an exports map with types and sideEffects: false.

Included from 0.2.0 (never published separately)

Parser rewrite: strict RFC 8866, a lossless model, typed WebRTC attributes. Breaking.

Fixed

• Time unit h in r=/t=/z= is 3600 s, not 86400 s (RFC 8866 §5.10).
• a=sendonly, a=recvonly and a=inactive were not recognized as directions
  (only sendrecv was).
• r=: active duration was read from the repeat-interval field, and offsets were dropped
  unless the line had more than three fields.
• b=: bwtype contained the b= prefix ("b=AS").
• u= overwrote i= instead of being stored as uri.
• Media-level i=, b= and k= overwrote the session-level fields.
• a=rtcp-fb kept only its first token (nack pli → nack).
• a=group with semantics other than the seven built-in ones was silently dropped.

Added

• SdpParseError (lineNumber, line, reason). It is thrown for malformed lines,
  unknown line types, line order violations (RFC 8866 §5), missing v=/o=/s=/t=
  and malformed typed attributes.
• z= (time zone adjustments), multiple b=/e=/p= lines, LF-only line endings.
• Typed attributes:
  a=candidate (RFC 8839), a=end-of-candidates (RFC 8840), a=ice-lite,
  a=ssrc / a=ssrc-group (RFC 5576), a=msid (RFC 8830), a=rid (RFC 8851),
  a=simulcast (RFC 8853), a=rtcp-mux (RFC 5761), a=rtcp-rsize (RFC 5506),
  a=extmap-allow-mixed (RFC 8285), a=rtcp (RFC 3605), a=sctp-port and
  a=max-message-size (RFC 8841). The a=extmap direction and extension attributes
  are now kept as well.
• A known attribute outside its scope (e.g. a=mid at session level), or rtpmap/fmtp/rtcp-fb
  for a payload type missing from m=, is kept in attributes instead of being lost.
• Test fixtures: real SDP from Chrome 156, Firefox 159, Chrome (2023) and SIP devices.

Changed (migration from 0.1.x)

0.1.x	0.2.0
protocolVersion?, origin?, sessionName?, timing?, media?	required (always present after parsing)
origin.sessVersion: number	string (like sessId; 64-bit values exist)
origin.netType: 'IN', addrType: 'IP4' | 'IP6' (others became undefined)	string, kept verbatim
repeat[] (session level)	timing[i].repeat[] (belongs to its t=)
emailAddress, phoneNumber	emailAddresses[], phoneNumbers[]
encryptionKeys[]	encryptionKey (at most one per level)
bandwidthInformation	bandwidths[]
ice.ufrag / ice.pwd / ice.options	iceUfrag / icePwd / iceOptions
dtls.algorithm + dtls.fingerprint	fingerprints[]: {hashFunction, fingerprint}
dtls.setup (defaulted to actpass when absent)	setup (absent when absent)
group[].token	group[].semantics
media[].ports: number[]	port, numberOfPorts?
media[].fmt[] (contentType: 'opus/48000/2', formatParameters)	codecs[] (encodingName, clockRate, encodingParameters, parameters, rtcpFeedback)
fmtp parameter without value → 'unknown'	null
media[].fmtp[] / media[].description (non-RTP)	formats[]
extmap: Record<number, string>	extmap: SdpExtmap[]
msid?: {streamId, trackId}	msid?: SdpMsid[]
attributes: Record<string, string | true | string[]>	attributes: {name, value?}[] in original order
types SdpAVMediaSection, SdpAppMediaSection, SdpOtherMediaSection	SdpMediaSection
types SdpAttribute, SdpIceAttributes, SdpDtlsAttribute	SdpGenericAttribute, flat fields
invalid input sometimes produced NaN or partial objects	SdpParseError

Upgrading from 0.1.x: the model changed, see the migration table above.

Install: npm i @irbisadm/simple-sdp@1.0.0