@blamejs/pki 0.1.21 → 0.1.23

package/lib/schema-pkix.js

@@ -2,15 +2,15 @@
 // Copyright (c) blamejs contributors
 "use strict";
 //
-// @internal — no operator-facing namespace. The documented surface is the
-// parsers that compose these factories (pki.schema.x509, pki.schema.crl, …).
+// @internal -- no operator-facing namespace. The documented surface is the
+// parsers that compose these factories (pki.schema.x509, pki.schema.crl, ...).
 //
 // Shared PKIX structure-schema factories (RFC 5280). Each is a namespace-
 // parameterized FACTORY: given an error namespace `ns` ({ prefix, E, oid }) it
 // returns an asn1-schema that walks the corresponding ASN.1 structure and emits
 // the caller's own <prefix>/* error codes. x509.js, crl.js, and future CMS/CSR
 // parsers compose these so AlgorithmIdentifier / Name / Extension are defined
-// once, not re-derived per format. This module is internal infrastructure — the
+// once, not re-derived per format. This module is internal infrastructure -- the
 // operator-facing surface is the parsers that consume it.
 
 var asn1 = require("./asn1-der");
@@ -30,8 +30,19 @@ var PEM_RE = /-----BEGIN ([A-Z0-9 ]+)-----([\s\S]*?)-----END \1-----/;
 
 // pemDecode(text, label, PemError): `label` (when truthy) is enforced, else the
 // first block is taken. Applies the LIMITS.PEM_MAX_BYTES cap before scanning.
+// The body must be CANONICAL base64 (RFC 4648 sec. 3.5): whole 4-character
+// groups, no stray padding, zero trailing bits in the final symbol. Node's
+// lenient Buffer.from would otherwise let several distinct PEM texts alias one
+// DER (PEM-layer malleability) and silently drop trailing content bits.
 function pemDecode(text, label, PemError) {
-  if (Buffer.isBuffer(text)) text = text.toString("latin1");
+  // The size cap is enforced on the raw byte length BEFORE the latin1 string
+  // copy: converting first would allocate a full-size string for an input the
+  // cap is about to reject, and a buffer above Node's max string length would
+  // escape as an untyped ERR_STRING_TOO_LONG instead of pem/too-large.
+  if (Buffer.isBuffer(text)) {
+    if (text.length > constants.LIMITS.PEM_MAX_BYTES) throw new PemError("pem/too-large", "PEM input exceeds size cap");
+    text = text.toString("latin1");
+  }
   if (typeof text !== "string") throw new PemError("pem/bad-input", "pemDecode expects a string or Buffer");
   if (text.length > constants.LIMITS.PEM_MAX_BYTES) throw new PemError("pem/too-large", "PEM input exceeds size cap");
   var m = PEM_RE.exec(text);
@@ -39,11 +50,21 @@ function pemDecode(text, label, PemError) {
   if (label && m[1] !== label) throw new PemError("pem/label-mismatch", "expected " + JSON.stringify(label) + " block, got " + JSON.stringify(m[1]));
   var b64 = m[2].replace(/[\r\n\t ]+/g, "");
   if (!/^[A-Za-z0-9+/]*={0,2}$/.test(b64)) throw new PemError("pem/bad-base64", "PEM body is not valid base64");
-  return Buffer.from(b64, "base64");
+  if (b64.length % 4 !== 0) throw new PemError("pem/bad-base64", "PEM base64 body must be whole 4-character groups (RFC 4648 sec. 3.5)");
+  var der = Buffer.from(b64, "base64");
+  if (der.toString("base64") !== b64) throw new PemError("pem/bad-base64", "PEM base64 body is not canonical (RFC 4648 sec. 3.5)");
+  return der;
 }
 
+// The label must be re-readable by this module's own PEM_RE (uppercase
+// alphanumeric words separated by single spaces, a subset of the RFC 7468
+// sec. 3 label grammar). Anything else (lowercase, a leading/trailing space,
+// an embedded "-----" or newline) would emit armor pemDecode itself rejects
+// or corrupt the armor grammar -- a config-time authoring fault, thrown here.
 function pemEncode(der, label, PemError) {
-  if (typeof label !== "string" || label.length === 0) throw new PemError("pem/bad-label", "pemEncode requires a label");
+  if (typeof label !== "string" || !/^[A-Z0-9]+( [A-Z0-9]+)*$/.test(label)) {
+    throw new PemError("pem/bad-label", "pemEncode requires an uppercase A-Z0-9 label with single spaces");
+  }
   var buf = Buffer.isBuffer(der) ? der : Buffer.from(der);
   var b64 = buf.toString("base64").replace(/(.{64})/g, "$1\n").replace(/\n$/, "");
   return "-----BEGIN " + label + "-----\n" + b64 + "\n-----END " + label + "-----\n";
@@ -63,8 +84,8 @@ function coerceToDer(input, opts) {
 
 // Does a Buffer carry PEM armor (a .pem read with fs.readFileSync) rather than
 // raw DER? It does iff "-----BEGIN" appears and everything before it is TEXT
-// (UTF-8 BOM / whitespace / RFC 7468 explanatory preamble). DER is binary — its
-// leading tag+length bytes are non-printable — so a non-SEQUENCE DER (a bare
+// (UTF-8 BOM / whitespace / RFC 7468 explanatory preamble). DER is binary -- its
+// leading tag+length bytes are non-printable -- so a non-SEQUENCE DER (a bare
 // SET / INTEGER) is NOT misrouted here; it decodes and fails closed structurally.
 function _isPemArmor(buf) {
   var head = buf.slice(0, 4096).toString("latin1");
@@ -83,8 +104,8 @@ function _isPemArmor(buf) {
 function decodeRoot(der, opts) {
   // A format whose wire encoding is normatively BER (RFC 7292 PKCS#12) opts
   // in to the codec's ber mode for the WHOLE decode. BER-vs-DER divergence is
-  // not always a decode-time throw — a definite-length constructed IMPLICIT
-  // OCTET STRING decodes strictly and only diverges at the typed reader — so
+  // not always a decode-time throw -- a definite-length constructed IMPLICIT
+  // OCTET STRING decodes strictly and only diverges at the typed reader -- so
   // a strict-first-retry-on-throw boundary would miss legal BER shapes. A
   // strict-DER input decodes identically under the ber mode, every other
   // strictness verdict included.
@@ -101,7 +122,7 @@ function runParse(input, opts) {
   return schema.walk(opts.topSchema, decodeRoot(coerceToDer(input, opts), opts), opts.ns).result;
 }
 
-// Distinguished-name attribute short labels (RFC 4514 §3 + common use).
+// Distinguished-name attribute short labels (RFC 4514 sec. 3 + common use).
 var DN_SHORT = {
   commonName:             "CN",
   countryName:            "C",
@@ -121,7 +142,7 @@ var DN_SHORT = {
 // The error/OID namespace every format module walks its schema under:
 // { prefix, E:(code,message,cause)=>new ErrorClass(...), oid }. Factored so a
 // format declares it in one line instead of repeating the error-constructor
-// closure (a caller that never passes a cause is unaffected — withCause ignores
+// closure (a caller that never passes a cause is unaffected -- withCause ignores
 // an undefined third arg).
 function makeNS(prefix, ErrorClass, oidModule) {
   return { prefix: prefix, E: function (code, message, cause) { return new ErrorClass(code, message, cause); }, oid: oidModule };
@@ -129,10 +150,10 @@ function makeNS(prefix, ErrorClass, oidModule) {
 
 // A bounded universal-INTEGER version reader. `accept` maps each legal wire value
 // (as a decimal string) to its surfaced version number; any other value is a
-// <prefix>/bad-version fault. The one genuine per-format divergence — the cert
+// <prefix>/bad-version fault. The one genuine per-format divergence -- the cert
 // rejects 0 and maps 1->2/2->3, a CRL accepts only 1->2, a CSR only 0->1, a PKCS#8
-// 0->1/1->2 — is expressed purely as the accept map (RFC 5280 §4.1.2.1 / §5.1.2.1,
-// RFC 2986 §4.1, RFC 5958 §2). read.integer is strict, so an ENUMERATED-tagged
+// 0->1/1->2 -- is expressed purely as the accept map (RFC 5280 sec. 4.1.2.1 / sec. 5.1.2.1,
+// RFC 2986 sec. 4.1, RFC 5958 sec. 2). read.integer is strict, so an ENUMERATED-tagged
 // version is rejected at the leaf (asn1/*).
 function versionReader(ns, accept) {
   return schema.decode(function (n) {
@@ -144,8 +165,20 @@ function versionReader(ns, accept) {
 
 // opts.implicitTag (optional): read the AlgorithmIdentifier as a [tag] IMPLICIT
 // SEQUENCE (a context-class constructed node whose children are algorithm + parameters),
-// for the RFC 5652 §6.2.4 pwri.keyDerivationAlgorithm [0]. With no opts the shape is a
+// for the RFC 5652 sec. 6.2.4 pwri.keyDerivationAlgorithm [0]. With no opts the shape is a
 // universal SEQUENCE, byte-identical to every existing caller.
+//
+// Parameter-rule tiering: only the params-MUST-BE-ABSENT families (ML-DSA,
+// SLH-DSA, the Edwards/Montgomery curves -- new registrations with no legacy
+// corpus) are enforced at parse time, in the build below. The CLASSICAL
+// per-OID rules (RSA parameters MUST be NULL, RFC 3279 sec. 2.2.1/2.3.1;
+// ecdsa-with-SHA2 parameters MUST be omitted, RFC 5758 sec. 3.2) are
+// deliberately NOT parse-time rejects: a long tail of deployed certificates
+// omits the RSA NULL, so the parse tier stays interop-lenient for those
+// families and the strict classical check runs on the signature-verification
+// path (path-validate's signature-algorithm registry), where the verdict
+// matters. SPKI key AlgorithmIdentifiers follow the same tiering: lenient at
+// parse, validated by the key-import layer that consumes them.
 function algorithmIdentifier(ns, opts) {
   opts = opts || {};
   return schema.seq([
@@ -156,14 +189,16 @@ function algorithmIdentifier(ns, opts) {
     arity: { min: 1 }, code: ns.prefix + "/bad-algorithm-identifier", what: "AlgorithmIdentifier",
     build: function (m, ctx) {
       var dotted = m.fields.algorithm.value;
-      // RFC 9909 §3 / RFC 9814 §4 / RFC 9881 §2 / RFC 8410 §3: for the PQC
-      // signature families (ML-DSA, SLH-DSA) and the Edwards/Montgomery curves,
-      // the `parameters` field MUST be absent — no explicit NULL, no bytes.
-      // Enforced ONCE here so every consumer of the shared AlgorithmIdentifier
-      // inherits the rule; the guard-parity bug class is structurally impossible.
+      // For every algorithm the registry marks parameters-absent -- the PQC
+      // signature families (ML-DSA RFC 9881, SLH-DSA RFC 9909/9814), the
+      // Edwards/Montgomery curves (RFC 8410), ML-KEM (RFC 9936), and the HKDF
+      // identifiers (RFC 8619) -- the `parameters` field MUST be absent: no
+      // explicit NULL, no bytes. Enforced ONCE here so every consumer of the
+      // shared AlgorithmIdentifier inherits the rule; the guard-parity bug
+      // class is structurally impossible.
       if (m.fields.parameters.present && ctx.oid.paramsMustBeAbsent(dotted)) {
         throw ctx.E(ctx.prefix + "/bad-algorithm-parameters",
-          "the " + (ctx.oid.name(dotted) || dotted) + " AlgorithmIdentifier parameters field MUST be absent (RFC 9909 §3 / RFC 9814 §4 / RFC 9881 §2 / RFC 8410 §3)");
+          "the " + (ctx.oid.name(dotted) || dotted) + " AlgorithmIdentifier parameters field MUST be absent");
       }
       return { oid: dotted, name: ctx.oid.name(dotted) || null, parameters: m.fields.parameters.present ? m.fields.parameters.node.bytes : null };
     },
@@ -173,44 +208,108 @@ function algorithmIdentifier(ns, opts) {
 // attrValueToString(ns): the AttributeValue decode-leaf. A malformed KNOWN
 // string type (invalid UTF-8, a non-IA5 byte, a PrintableString character
 // outside its set, ...) surfaces as an asn1/bad-* content error and must fail
-// closed — do NOT hex-encode it away, or the decoder's strict string validation
+// closed -- do NOT hex-encode it away, or the decoder's strict string validation
 // is silently bypassed on the DN path. A value that is simply not a decodable
 // primitive string is NOT malformed and stays representable: an ANY-typed
 // non-string tag (asn1/expected-string) or a constructed universal type such as
-// a SEQUENCE (asn1/expected-primitive) renders per RFC 4514 §2.4 as "#" plus the
+// a SEQUENCE (asn1/expected-primitive) renders per RFC 4514 sec. 2.4 as "#" plus the
 // hex of its FULL DER encoding (node.bytes), round-tripping intact.
+// A GENUINE string whose first character is '#' or '\' is surfaced with a
+// leading '\' escape (RFC 4514 sec. 2.4) so it can never collide with the
+// '#'+hex form; the encode direction strips that one leading escape back off.
 function attrValueToString(ns) {
   return schema.decode(function (node) {
-    try { return asn1.read.string(node); }
+    var s;
+    try { s = asn1.read.string(node); }
     catch (e) {
       if (!e || (e.code !== "asn1/expected-string" && e.code !== "asn1/expected-primitive")) {
         throw ns.E(ns.prefix + "/bad-atv", "malformed string in attribute value: " + ((e && e.message) || String(e)));
       }
       return "#" + node.bytes.toString("hex");
     }
+    if (s.charAt(0) === "#" || s.charAt(0) === "\\") return "\\" + s;
+    return s;
   }, function (value) {
-    // encode: a "#hex" escape (RFC 4514 §2.4) round-trips its raw DER verbatim; any
-    // other string encodes as UTF8String (the decode does not preserve the original
+    // encode: a leading '\' escapes the next character (the decode's escape of a
+    // literal '#'/'\'); a "#hex" form (RFC 4514 sec. 2.4) round-trips its raw DER
+    // verbatim -- the hex is validated (even-length hex digits encoding exactly one
+    // DER TLV) and throws rather than silently emitting truncated bytes, since
+    // Buffer.from(str, "hex") stops at the first invalid character. Any other
+    // string encodes as UTF8String (the decode does not preserve the original
     // string type, so this is the canonical re-encoding, not a byte-exact one).
-    if (typeof value === "string" && value.charAt(0) === "#") return Buffer.from(value.slice(1), "hex");
+    if (typeof value === "string" && value.charAt(0) === "\\") return asn1.build.utf8(value.slice(1));
+    if (typeof value === "string" && value.charAt(0) === "#") {
+      var hex = value.slice(1);
+      if (!/^(?:[0-9A-Fa-f]{2})+$/.test(hex)) {
+        throw ns.E(ns.prefix + "/bad-atv", "a #hex attribute value must be a non-empty even run of hex digits (RFC 4514 sec. 2.4)");
+      }
+      var raw = Buffer.from(hex, "hex");
+      try { asn1.decode(raw); }
+      catch (e) { throw ns.E(ns.prefix + "/bad-atv", "a #hex attribute value must encode exactly one DER TLV", e); }
+      return raw;
+    }
     return asn1.build.utf8(value);
   });
 }
 
+// The always-escape specials of RFC 4514 sec. 2.4, plus its positional rules:
+// NUL escapes as \00, a trailing space as '\ ', and a leading '#' or space as
+// '\#' / '\ ' -- without them the rendered dn is ambiguous (a literal '#0500'
+// value would be indistinguishable from the hexstring form) and leading /
+// trailing spaces would be silently significant.
 function _escapeDnValue(v) {
-  return v.replace(/([,+"\\<>;])/g, "\\$1");
+  var s = v.replace(/([,+"\\<>;])/g, "\\$1").split("\u0000").join("\\00");
+  if (s.length && s.charAt(s.length - 1) === " ") s = s.slice(0, -1) + "\\ ";
+  if (s.charAt(0) === "#" || s.charAt(0) === " ") s = "\\" + s;
+  return s;
+}
+
+// The dn RENDERING of a surfaced AttributeValue: the '#'-leading hex form (a
+// non-string value) renders bare -- escaping its '#' would turn the RFC 4514
+// hexstring form into an escaped literal -- while a genuine string (surfaced
+// with its leading '#'/'\' escape, see attrValueToString) is unescaped back to
+// the semantic string and then rendered with the full escape set.
+function _dnDisplayValue(v) {
+  if (typeof v !== "string") return v;
+  if (v.charAt(0) === "#") return v;
+  return _escapeDnValue(v.charAt(0) === "\\" ? v.slice(1) : v);
+}
+
+// time(ns): the RFC 5280 Time production under its profile encoding rule
+// (sec. 4.1.2.5 for certificate validity, restated for the CRL fields in
+// sec. 5.1.2.4-5.1.2.6): a date through the year 2049 MUST be UTCTime and
+// GeneralizedTime is reserved for 2050 onward, so a GeneralizedTime carrying a
+// 1950..2049 date is a non-conforming second encoding of a representable value
+// and is rejected (<prefix>/bad-time) -- accepting it would let two wire forms
+// alias one abstract time (a parser-differential channel). The engine's
+// schema.time already applies the cutover on ENCODE; this leaf composes it and
+// enforces the same rule on DECODE. A GeneralizedTime-only field (an RFC 3161
+// genTime, a CRL invalidityDate, an RFC 5755 AttCertValidityPeriod) must NOT
+// take this leaf -- its syntax has no UTCTime alternative to cut over to.
+function time(ns) {
+  var base = schema.time(ns);
+  return schema.decode(function (n, ctx) {
+    var d = base.fn(n, ctx);
+    if (n.tagNumber === asn1.TAGS.GENERALIZED_TIME) {
+      var y = d.getUTCFullYear();
+      if (y >= 1950 && y < 2050) {
+        throw ctx.E(ctx.prefix + "/bad-time", "a date through 2049 must be encoded as UTCTime, not GeneralizedTime (RFC 5280 sec. 4.1.2.5)");
+      }
+    }
+    return d;
+  }, base.write);
 }
 
 // Name ::= RDNSequence ::= SEQUENCE OF RelativeDistinguishedName; RDN ::= SET OF
 // AttributeTypeAndValue ::= SEQUENCE { type OID, value ANY }. The atv asserts
-// bare-constructed (min 2) — matching the historical guard that never checked
-// the SEQUENCE tag — and repeated RDN attribute types stay legal (no uniqueness).
+// its SEQUENCE tag (a SET-tagged body is a different ASN.1 type), and repeated
+// RDN attribute types stay legal (no uniqueness).
 function attributeTypeAndValue(ns) {
   return schema.seq([
     schema.field("type", schema.oidLeaf()),
     schema.field("value", attrValueToString(ns)),
   ], {
-    assert: "constructed", arity: { min: 2 }, code: ns.prefix + "/bad-atv", what: "AttributeTypeAndValue",
+    assert: "sequence", arity: { min: 2 }, code: ns.prefix + "/bad-atv", what: "AttributeTypeAndValue",
     build: function (m, ctx) {
       var typeOid = m.fields.type.value;
       return { type: typeOid, name: ctx.oid.name(typeOid) || null, value: m.fields.value.value };
@@ -218,13 +317,13 @@ function attributeTypeAndValue(ns) {
   });
 }
 function relativeDistinguishedName(ns) {
-  // RelativeDistinguishedName ::= SET SIZE (1..MAX) — an empty SET {} is malformed.
+  // RelativeDistinguishedName ::= SET SIZE (1..MAX) -- an empty SET {} is malformed.
   return schema.setOf(attributeTypeAndValue(ns), { assert: "set", min: 1, code: ns.prefix + "/bad-rdn", what: "RelativeDistinguishedName" });
 }
-// opts.implicitTag (optional): read the Name as a [tag] IMPLICIT RDNSequence — the
+// opts.implicitTag (optional): read the Name as a [tag] IMPLICIT RDNSequence -- the
 // context tag REPLACES the universal SEQUENCE tag, so the node is a context-class
 // constructed [tag] whose children ARE the RDN SETs (the CRMF CertTemplate issuer
-// [3] / subject [5] IMPLICIT arm, RFC 4211 §5). With no opts the shape is a bare
+// [3] / subject [5] IMPLICIT arm, RFC 4211 sec. 5). With no opts the shape is a bare
 // universal SEQUENCE OF, byte-identical to every existing caller.
 function name(ns, opts) {
   opts = opts || {};
@@ -236,7 +335,7 @@ function name(ns, opts) {
         var a = atvItem.value.result; // atvItem.value = the atv seq-match; .result = its build result
         atvs.push(a);
         var label = (a.name && DN_SHORT[a.name]) || a.name || a.type;
-        atvParts.push(label + "=" + _escapeDnValue(a.value));
+        atvParts.push(label + "=" + _dnDisplayValue(a.value));
       });
       rdns.push(atvs);
       parts.push(atvParts.join("+"));
@@ -251,25 +350,85 @@ function name(ns, opts) {
     assert: "sequence", code: ns.prefix + "/bad-name", what: "Name", build: nameBuild });
 }
 
-// GeneralName ::= CHOICE (RFC 5280 §4.2.1.6). A validate-and-surface-raw leaf for a
+// GeneralName ::= CHOICE (RFC 5280 sec. 4.2.1.6). A validate-and-surface-raw leaf for a
 // caller that keeps the value RAW but needs it to be a well-formed GeneralName: it
 // checks the chosen alternative's tag, form (constructed vs primitive per X.690
-// §10.2), and content — otherName [0] is a SEQUENCE { type-id OID, value [0] EXPLICIT
+// sec. 10.2), and content -- otherName [0] is a SEQUENCE { type-id OID, value [0] EXPLICIT
 // }; x400Address [3] / ediPartyName [5] are non-empty constructed; directoryName [4]
 // EXPLICIT wraps a valid Name; rfc822Name [1] / dNSName [2] / uniformResourceIdentifier
 // [6] are primitive non-empty IA5String (7-bit); iPAddress [7] is a 4- or 16-octet
-// OCTET STRING; registeredID [8] is a primitive OBJECT IDENTIFIER — then surfaces the
+// OCTET STRING; registeredID [8] is a primitive OBJECT IDENTIFIER -- then surfaces the
 // value RAW ({ bytes, tagClass, tagNumber }). `opts.code` is the caller's error code
 // (e.g. ocsp/bad-requestor-name, tsp/bad-tsa). Shared so the two parsers cannot drift.
 // opts.decodeValue (default false): in addition to the raw { bytes, tagClass,
-//   tagNumber }, surface the DECODED `value` per arm — IA5 text (string), an
+//   tagNumber }, surface the DECODED `value` per arm -- IA5 text (string), an
 //   iPAddress Buffer, a directoryName { rdns, dn }, a registeredID OID string,
 //   or an otherName { typeId, valueBytes }. The path validator's name-constraint
 //   matcher needs the decoded value; the tsp/ocsp/attrcert consumers pass no
 //   flag and get the byte-identical raw-only shape.
 // opts.subtreeBase (default false): the GeneralSubtree.base form (RFC 5280
-//   §4.2.1.10) — an iPAddress base is an address+mask (8 octets IPv4 / 32 IPv6),
+//   sec. 4.2.1.10) -- an iPAddress base is an address+mask (8 octets IPv4 / 32 IPv6),
 //   NOT the 4/16-octet SAN address form. Only the iPAddress size rule changes.
+// A GeneralizedTime-only leaf. Several formats fix their times to
+// GeneralizedTime and reject UTCTime (OCSP RFC 6960, TSP RFC 3161 sec. 2.4.2,
+// CMP RFC 9810 sec. 5.1.1, attribute-certificate validity RFC 5755 sec. 4.2.6);
+// ONE ns-parameterized factory owns the tag assert so the per-format code and
+// message differ but the check cannot drift. opts.allowFractional enables the
+// X.690 sec. 11.7 fractional-seconds profile (the RFC 3161 genTime).
+function generalizedTime(ns, opts) {
+  opts = opts || {};
+  var code = opts.code || (ns.prefix + "/bad-time");
+  var message = opts.message || "the time must be a GeneralizedTime";
+  var readOpts = opts.allowFractional ? { allowFractional: true } : undefined;
+  return schema.decode(function (n, ctx) {
+    if (n.tagClass !== "universal" || n.tagNumber !== asn1.TAGS.GENERALIZED_TIME) {
+      throw ctx.E(code, message);
+    }
+    return asn1.read.time(n, readOpts);
+  });
+}
+
+// A UTF8String-only leaf -- the PKIFreeText element shape (RFC 3161 sec. 2.4.2,
+// RFC 9810 sec. 5.1.1) shared by the TSP and CMP parsers.
+function utf8Text(ns, opts) {
+  opts = opts || {};
+  var code = opts.code || (ns.prefix + "/bad-freetext");
+  var message = opts.message || "the element must be a UTF8String";
+  return schema.decode(function (n, ctx) {
+    if (n.tagClass !== "universal" || n.tagNumber !== asn1.TAGS.UTF8_STRING) {
+      throw ctx.E(code, message);
+    }
+    return asn1.read.string(n);
+  });
+}
+
+// A raw NON-EMPTY universal SEQUENCE leaf -- surfaces the exact TLV bytes of an
+// embedded element (a Certificate, a CertificateList, a PKIPublicationInfo)
+// after asserting the SEQUENCE tag and at least one child. An empty SEQUENCE
+// (30 00) has the right tag but is none of those structures, so it rejects
+// rather than surfacing degenerate bytes to certificate processing.
+function rawNonEmptySequence(ns, opts) {
+  opts = opts || {};
+  var code = opts.code || (ns.prefix + "/bad-sequence");
+  var message = opts.message || "expected a non-empty universal SEQUENCE";
+  return schema.decode(function (n, ctx) {
+    if (!(n.tagClass === "universal" && n.tagNumber === asn1.TAGS.SEQUENCE && n.children && n.children.length >= 1)) {
+      throw ctx.E(code, message);
+    }
+    return n.bytes;
+  });
+}
+
+// CRLReason ::= ENUMERATED (RFC 5280 sec. 5.3.1) -- value 7 is unused/reserved.
+// The ONE canonical value->name table every consumer validates against: the
+// CRL reasonCode decoder surfaces the numeric code, the OCSP RevokedInfo
+// decoder the name; both draw the legal set from here so they cannot drift.
+var CRL_REASON_NAMES = {
+  "0": "unspecified", "1": "keyCompromise", "2": "cACompromise", "3": "affiliationChanged",
+  "4": "superseded", "5": "cessationOfOperation", "6": "certificateHold",
+  "8": "removeFromCRL", "9": "privilegeWithdrawn", "10": "aACompromise",
+};
+
 var GN_CONSTRUCTED = { 0: 1, 3: 1, 4: 1, 5: 1 };
 var GN_IA5 = { 1: 1, 2: 1, 6: 1 };
 function generalName(ns, opts) {
@@ -280,13 +439,13 @@ function generalName(ns, opts) {
   var NAME = name(ns);
   return schema.decode(function (n, ctx) {
     if (n.tagClass !== "context" || n.tagNumber < 0 || n.tagNumber > 8) {
-      throw ctx.E(code, "value must be a GeneralName (context tag [0]..[8]) (RFC 5280 §4.2.1.6)");
+      throw ctx.E(code, "value must be a GeneralName (context tag [0]..[8]) (RFC 5280 sec. 4.2.1.6)");
     }
     var t = n.tagNumber;
     var constructed = !!n.children;
     var value;
     if (GN_CONSTRUCTED[t]) {
-      if (!constructed || n.children.length < 1) throw ctx.E(code, "GeneralName [" + t + "] must be a non-empty constructed value (RFC 5280 §4.2.1.6)");
+      if (!constructed || n.children.length < 1) throw ctx.E(code, "GeneralName [" + t + "] must be a non-empty constructed value (RFC 5280 sec. 4.2.1.6)");
       if (t === 0) {
         // otherName ::= SEQUENCE { type-id OBJECT IDENTIFIER, value [0] EXPLICIT ANY }.
         if (n.children.length !== 2) throw ctx.E(code, "GeneralName otherName [0] must be a SEQUENCE { type-id, value [0] }");
@@ -299,17 +458,17 @@ function generalName(ns, opts) {
         }
         if (decodeValue) value = { typeId: typeId, valueBytes: ov.children[0].bytes };
       } else if (t === 4) {
-        // directoryName [4] EXPLICIT Name — validate the wrapped RDNSequence.
+        // directoryName [4] EXPLICIT Name -- validate the wrapped RDNSequence.
         if (n.children.length !== 1) throw ctx.E(code, "GeneralName directoryName [4] must wrap exactly one Name");
         var dnMatch = schema.walk(NAME, n.children[0], ctx);
         if (decodeValue) value = dnMatch.result;
       }
     } else {
-      if (constructed) throw ctx.E(code, "GeneralName [" + t + "] must be primitive (X.690 §10.2)");
+      if (constructed) throw ctx.E(code, "GeneralName [" + t + "] must be primitive (X.690 sec. 10.2)");
       if (GN_IA5[t]) {
         if (n.content.length === 0) throw ctx.E(code, "GeneralName [" + t + "] must be a non-empty IA5String");
         for (var i = 0; i < n.content.length; i++) {
-          // 7-bit IA5, and no C0/DEL control byte — an embedded NUL/control in a
+          // 7-bit IA5, and no C0/DEL control byte -- an embedded NUL/control in a
           // dNSName/rfc822Name/URI enables a name-truncation/confusion bypass
           // (CVE-2009-2408 class) downstream, so reject it at decode.
           if (n.content[i] < 0x20 || n.content[i] > 0x7e) throw ctx.E(code, "GeneralName [" + t + "] must be a printable IA5String (no control bytes)");
@@ -317,7 +476,7 @@ function generalName(ns, opts) {
         if (decodeValue) value = n.content.toString("latin1");
       } else if (t === 7) {
         if (subtreeBase) {
-          if (n.content.length !== 8 && n.content.length !== 32) throw ctx.E(code, "GeneralName iPAddress [7] subtree base must be an 8- or 32-octet address+mask (RFC 5280 §4.2.1.10)");
+          if (n.content.length !== 8 && n.content.length !== 32) throw ctx.E(code, "GeneralName iPAddress [7] subtree base must be an 8- or 32-octet address+mask (RFC 5280 sec. 4.2.1.10)");
         } else if (n.content.length !== 4 && n.content.length !== 16) {
           throw ctx.E(code, "GeneralName iPAddress [7] must be a 4- or 16-octet address");
         }
@@ -337,14 +496,14 @@ function generalName(ns, opts) {
   });
 }
 
-// GeneralNames ::= SEQUENCE OF GeneralName (RFC 5280 §4.2.1.6), SIZE (1..MAX). Every
+// GeneralNames ::= SEQUENCE OF GeneralName (RFC 5280 sec. 4.2.1.6), SIZE (1..MAX). Every
 // element is validated by generalName (its CHOICE alternative's form + content) and
 // surfaced raw, so a caller carrying a GeneralNames field cannot accept a malformed
 // element by treating the whole sequence as opaque bytes. Returns { names, bytes }
-// where each `names[i]` is the generalName leaf's return — { bytes, tagClass,
-// tagNumber } and, when opts.decodeValue is set, the decoded `value` — and `bytes`
+// where each `names[i]` is the generalName leaf's return -- { bytes, tagClass,
+// tagNumber } and, when opts.decodeValue is set, the decoded `value` -- and `bytes`
 // is the raw outer DER.
-// `opts.implicitTag` handles a [tag] IMPLICIT GeneralNames — the context tag REPLACES
+// `opts.implicitTag` handles a [tag] IMPLICIT GeneralNames -- the context tag REPLACES
 // the universal SEQUENCE tag (RFC 5755 Holder.entityName [1]); otherwise it is a bare
 // universal SEQUENCE OF. `opts.code` is the caller's error code. Shared so the x509 /
 // attribute-certificate / (future) CRMF parsers validate a GeneralNames identically.
@@ -361,15 +520,37 @@ function generalNames(ns, opts) {
   return schema.seqOf(gn, { assert: "sequence", min: 1, code: code, what: opts.what || "GeneralNames", build: build });
 }
 
-// certExtensionDecoders(ns) — the ns-parameterized RFC 5280 §4.2.1 extension
+// certExtensionDecoders(ns) -- the ns-parameterized RFC 5280 sec. 4.2.1 extension
 // VALUE decoders. `x509.parse` surfaces each extension as { oid, name, critical,
 // value } with `value` the raw inner OCTET-STRING content (a Buffer); the path
 // validator and the future complete-extension-set surface need the STRUCTURE
 // inside. Each decoder takes that raw Buffer and returns the decoded value, or
 // throws a typed `<prefix>/bad-*` code, mirroring the CRL's fail-closed
-// `decodeExt` pattern (asn1.decode → read.* / schema.walk, wrapped). Returns
+// `decodeExt` pattern (asn1.decode -> read.* / schema.walk, wrapped). Returns
 // { byOid } keyed by dotted OID so a consumer dispatches by ext.oid.
 var _T = asn1.TAGS;
+
+// assertPolicyQualifiers(qNode, fail) -- RFC 5280 sec. 4.2.1.4: a PolicyInformation's
+// policyQualifiers is a non-empty SEQUENCE OF PolicyQualifierInfo, each a SEQUENCE
+// { policyQualifierId OID, qualifier } of EXACTLY two members led by an OID. The
+// structure is validated fail-closed; the qualifier body stays opaque (surfaced
+// raw by the caller). `fail(msg)` throws the caller's typed code. Shared by the
+// certificatePolicies extension decoder and the RFC 5035 ESS SigningCertificate
+// policies field so the same PolicyInformation shape cannot drift between them.
+// `fail(msg, cause?)` throws the caller's typed code, carrying the underlying
+// leaf fault as the error cause when one is available.
+function assertPolicyQualifiers(qNode, fail) {
+  if (qNode.tagClass !== "universal" || qNode.tagNumber !== _T.SEQUENCE || !qNode.children || !qNode.children.length) {
+    fail("policyQualifiers must be a non-empty SEQUENCE (RFC 5280 sec. 4.2.1.4)");
+  }
+  qNode.children.forEach(function (pq) {
+    if (pq.tagClass !== "universal" || pq.tagNumber !== _T.SEQUENCE || !pq.children || pq.children.length !== 2) {
+      fail("policyQualifiers element must be a PolicyQualifierInfo SEQUENCE { policyQualifierId, qualifier } (RFC 5280 sec. 4.2.1.4)");
+    }
+    try { asn1.read.oid(pq.children[0]); } catch (e) { fail("PolicyQualifierInfo must lead with a policyQualifierId OID", e); }
+  });
+}
+
 function certExtensionDecoders(ns) {
   var GN_SUBTREE = generalName(ns, { decodeValue: true, subtreeBase: true, code: ns.prefix + "/bad-name-constraints" });
 
@@ -382,7 +563,7 @@ function certExtensionDecoders(ns) {
   function seqChildren(buf, code, what) {
     var n = decodeTop(buf, code, what);
     if (n.tagClass !== "universal" || n.tagNumber !== _T.SEQUENCE || !n.children) {
-      throw ns.E(code, what + " must be a SEQUENCE (RFC 5280 §4.2.1)");
+      throw ns.E(code, what + " must be a SEQUENCE (RFC 5280 sec. 4.2.1)");
     }
     return n.children;
   }
@@ -399,13 +580,13 @@ function certExtensionDecoders(ns) {
     if (kids[i] && kids[i].tagClass === "universal" && kids[i].tagNumber === _T.BOOLEAN) {
       var v;
       try { v = asn1.read.boolean(kids[i]); } catch (e) { throw ns.E(C, "BasicConstraints cA must be a BOOLEAN", e); }
-      if (v !== true) throw ns.E(C, "BasicConstraints cA DEFAULT FALSE must be omitted, not an explicit FALSE (X.690 §11.5)");
+      if (v !== true) throw ns.E(C, "BasicConstraints cA DEFAULT FALSE must be omitted, not an explicit FALSE (X.690 sec. 11.5)");
       cA = true; i++;
     }
     if (kids[i] && kids[i].tagClass === "universal" && kids[i].tagNumber === _T.INTEGER) {
-      if (!cA) throw ns.E(C, "BasicConstraints pathLenConstraint is only permitted when cA is TRUE (RFC 5280 §4.2.1.9)");
+      if (!cA) throw ns.E(C, "BasicConstraints pathLenConstraint is only permitted when cA is TRUE (RFC 5280 sec. 4.2.1.9)");
       var pl = readInt(kids[i], C, "pathLenConstraint");
-      // Bounded before Number() narrowing — a value past the safe-integer range
+      // Bounded before Number() narrowing -- a value past the safe-integer range
       // would round silently and be compared as a path-length ceiling, so it is
       // exact-or-rejected (the same rule the RSASSA-PSS/PKCS#12 counters follow).
       if (pl < 0n || pl > 2147483647n) throw ns.E(C, "BasicConstraints pathLenConstraint must be a non-negative integer within range");
@@ -421,15 +602,15 @@ function certExtensionDecoders(ns) {
   function keyUsage(buf) {
     var C = ns.prefix + "/bad-key-usage";
     var n = decodeTop(buf, C, "KeyUsage");
-    if (n.tagClass !== "universal" || n.tagNumber !== _T.BIT_STRING) throw ns.E(C, "KeyUsage must be a BIT STRING (RFC 5280 §4.2.1.3)");
+    if (n.tagClass !== "universal" || n.tagNumber !== _T.BIT_STRING) throw ns.E(C, "KeyUsage must be a BIT STRING (RFC 5280 sec. 4.2.1.3)");
     var bs;
     try { bs = asn1.read.bitString(n); } catch (e) { throw ns.E(C, "KeyUsage must be a well-formed BIT STRING", e); }
-    // At least one bit MUST be set (RFC 5280 §4.2.1.3) — an all-zero value (a
+    // At least one bit MUST be set (RFC 5280 sec. 4.2.1.3) -- an all-zero value (a
     // non-empty byte run of only zero bits, e.g. 03 02 07 00) is malformed too.
     var anyBit = false;
     for (var z = 0; z < bs.bytes.length; z++) { if (bs.bytes[z] !== 0) { anyBit = true; break; } }
-    if (!anyBit) throw ns.E(C, "KeyUsage must assert at least one bit (RFC 5280 §4.2.1.3)");
-    // KeyUsage is a NamedBitList — enforce the shared X.690 §11.2.2 minimal-DER rule.
+    if (!anyBit) throw ns.E(C, "KeyUsage must assert at least one bit (RFC 5280 sec. 4.2.1.3)");
+    // KeyUsage is a NamedBitList -- enforce the shared X.690 sec. 11.2.2 minimal-DER rule.
     schema.assertMinimalNamedBits(bs.unusedBits, bs.bytes, function (m) { throw ns.E(C, m); });
     var out = {};
     KU_BITS.forEach(function (nm, bit) {
@@ -442,9 +623,9 @@ function certExtensionDecoders(ns) {
   // GeneralSubtree ::= SEQUENCE { base GeneralName, minimum [0] DEFAULT 0, maximum [1] OPTIONAL }
   function subtreeList(node, C) {
     // node is the [0]/[1] IMPLICIT SEQUENCE OF GeneralSubtree (context-constructed).
-    // GeneralSubtrees is SIZE(1..MAX) — an explicitly present but empty subtree
+    // GeneralSubtrees is SIZE(1..MAX) -- an explicitly present but empty subtree
     // list (a0 00) absorbs no constraint and is malformed, not a no-op.
-    if (!node.children || node.children.length < 1) throw ns.E(C, "NameConstraints permittedSubtrees/excludedSubtrees must be a non-empty GeneralSubtrees (SIZE 1..MAX, RFC 5280 §4.2.1.10)");
+    if (!node.children || node.children.length < 1) throw ns.E(C, "NameConstraints permittedSubtrees/excludedSubtrees must be a non-empty GeneralSubtrees (SIZE 1..MAX, RFC 5280 sec. 4.2.1.10)");
     return node.children.map(function (st) {
       if (st.tagClass !== "universal" || st.tagNumber !== _T.SEQUENCE || !st.children || st.children.length < 1) {
         throw ns.E(C, "GeneralSubtree must be a SEQUENCE { base, minimum?, maximum? }");
@@ -453,8 +634,8 @@ function certExtensionDecoders(ns) {
       for (var j = 1; j < st.children.length; j++) {
         var f = st.children[j];
         if (f.tagClass !== "context" || (f.tagNumber !== 0 && f.tagNumber !== 1)) throw ns.E(C, "GeneralSubtree has an unexpected field");
-        if (f.tagNumber === 0) throw ns.E(C, "GeneralSubtree minimum DEFAULT 0 must be omitted (RFC 5280 §4.2.1.10 requires minimum = 0)");
-        if (f.tagNumber === 1) throw ns.E(C, "GeneralSubtree maximum is not permitted in the RFC 5280 profile (§4.2.1.10)");
+        if (f.tagNumber === 0) throw ns.E(C, "GeneralSubtree minimum DEFAULT 0 must be omitted (RFC 5280 sec. 4.2.1.10 requires minimum = 0)");
+        if (f.tagNumber === 1) throw ns.E(C, "GeneralSubtree maximum is not permitted in the RFC 5280 profile (sec. 4.2.1.10)");
       }
       return { base: base };
     });
@@ -471,7 +652,7 @@ function certExtensionDecoders(ns) {
       else if (k.tagNumber === 1) { if (sawE) throw ns.E(C, "duplicate excludedSubtrees"); sawE = true; excluded = subtreeList(k, C); }
       else throw ns.E(C, "NameConstraints has an unexpected field [" + k.tagNumber + "]");
     });
-    if (!sawP && !sawE) throw ns.E(C, "NameConstraints must contain permittedSubtrees or excludedSubtrees (RFC 5280 §4.2.1.10)");
+    if (!sawP && !sawE) throw ns.E(C, "NameConstraints must contain permittedSubtrees or excludedSubtrees (RFC 5280 sec. 4.2.1.10)");
     return { permittedSubtrees: permitted, excludedSubtrees: excluded };
   }
 
@@ -479,36 +660,25 @@ function certExtensionDecoders(ns) {
   function certificatePolicies(buf) {
     var C = ns.prefix + "/bad-policy";
     var kids = seqChildren(buf, C, "CertificatePolicies");
-    if (kids.length < 1) throw ns.E(C, "CertificatePolicies must contain at least one PolicyInformation (RFC 5280 §4.2.1.4)");
+    if (kids.length < 1) throw ns.E(C, "CertificatePolicies must contain at least one PolicyInformation (RFC 5280 sec. 4.2.1.4)");
     var seen = {};
     return kids.map(function (pi) {
       // PolicyInformation ::= SEQUENCE { policyIdentifier, policyQualifiers
-      // SEQUENCE SIZE(1..MAX) OPTIONAL } — exactly one or two fields; a second
+      // SEQUENCE SIZE(1..MAX) OPTIONAL } -- exactly one or two fields; a second
       // field, if present, MUST be a SEQUENCE. Extra/mis-typed fields are malformed.
       if (pi.tagClass !== "universal" || pi.tagNumber !== _T.SEQUENCE || !pi.children || pi.children.length < 1 || pi.children.length > 2) {
         throw ns.E(C, "PolicyInformation must be a SEQUENCE { policyIdentifier, policyQualifiers? }");
       }
       var pid;
       try { pid = asn1.read.oid(pi.children[0]); } catch (e) { throw ns.E(C, "PolicyInformation policyIdentifier must be an OBJECT IDENTIFIER", e); }
-      if (seen[pid]) throw ns.E(C, "duplicate policy OID " + pid + " (RFC 5280 §4.2.1.4)");
+      if (seen[pid]) throw ns.E(C, "duplicate policy OID " + pid + " (RFC 5280 sec. 4.2.1.4)");
       seen[pid] = true;
       var qualifiers = null;
       if (pi.children.length > 1) {
         var q = pi.children[1];
-        if (q.tagClass !== "universal" || q.tagNumber !== _T.SEQUENCE || !q.children || !q.children.length) {
-          throw ns.E(C, "PolicyInformation policyQualifiers must be a non-empty SEQUENCE (RFC 5280 §4.2.1.4)");
-        }
-        // Each element is a PolicyQualifierInfo ::= SEQUENCE { policyQualifierId
-        // OID, qualifier ANY DEFINED BY policyQualifierId } — EXACTLY two members
-        // (qualifier is not OPTIONAL). A SEQUENCE missing the qualifier or
-        // carrying trailing extra fields is malformed; the qualifier body itself
-        // stays opaque.
-        q.children.forEach(function (pq) {
-          if (pq.tagClass !== "universal" || pq.tagNumber !== _T.SEQUENCE || !pq.children || pq.children.length !== 2) {
-            throw ns.E(C, "policyQualifiers element must be a PolicyQualifierInfo SEQUENCE { policyQualifierId, qualifier } (RFC 5280 §4.2.1.4)");
-          }
-          try { asn1.read.oid(pq.children[0]); } catch (e) { throw ns.E(C, "PolicyQualifierInfo must lead with a policyQualifierId OID", e); }
-        });
+        // PolicyQualifierInfo structure validated by the shared assertion; the
+        // qualifier body stays opaque (surfaced raw).
+        assertPolicyQualifiers(q, function (msg, cause) { throw ns.E(C, msg, cause); });
         qualifiers = q.bytes;
       }
       return { policyIdentifier: pid, qualifiersBytes: qualifiers };
@@ -519,7 +689,7 @@ function certExtensionDecoders(ns) {
   function policyMappings(buf) {
     var C = ns.prefix + "/bad-policy";
     var kids = seqChildren(buf, C, "PolicyMappings");
-    if (kids.length < 1) throw ns.E(C, "PolicyMappings must contain at least one mapping (RFC 5280 §4.2.1.5)");
+    if (kids.length < 1) throw ns.E(C, "PolicyMappings must contain at least one mapping (RFC 5280 sec. 4.2.1.5)");
     return kids.map(function (mp) {
       if (mp.tagClass !== "universal" || mp.tagNumber !== _T.SEQUENCE || !mp.children || mp.children.length !== 2) {
         throw ns.E(C, "policy mapping must be a SEQUENCE { issuerDomainPolicy, subjectDomainPolicy }");
@@ -535,7 +705,7 @@ function certExtensionDecoders(ns) {
   function policyConstraints(buf) {
     var C = ns.prefix + "/bad-policy";
     var kids = seqChildren(buf, C, "PolicyConstraints");
-    if (kids.length < 1) throw ns.E(C, "PolicyConstraints must contain at least one field (RFC 5280 §4.2.1.11)");
+    if (kids.length < 1) throw ns.E(C, "PolicyConstraints must contain at least one field (RFC 5280 sec. 4.2.1.11)");
     var rep = null, ipm = null, pcLastTag = -1;
     kids.forEach(function (k) {
       if (k.tagClass !== "context") throw ns.E(C, "PolicyConstraints fields are context-tagged [0]/[1]");
@@ -544,7 +714,7 @@ function certExtensionDecoders(ns) {
       var v;
       try { v = asn1.read.integerImplicit(k, k.tagNumber); } catch (e) { throw ns.E(C, "PolicyConstraints field must be an INTEGER", e); }
       // Bounded before Number() narrowing (a skip count past the safe-integer
-      // range would round silently) — exact-or-rejected.
+      // range would round silently) -- exact-or-rejected.
       if (v < 0n || v > 2147483647n) throw ns.E(C, "PolicyConstraints skip count must be a non-negative integer within range");
       if (k.tagNumber === 0) rep = Number(v);
       else if (k.tagNumber === 1) ipm = Number(v);
@@ -557,15 +727,15 @@ function certExtensionDecoders(ns) {
   function inhibitAnyPolicy(buf) {
     var C = ns.prefix + "/bad-policy";
     var n = decodeTop(buf, C, "InhibitAnyPolicy");
-    if (n.tagClass !== "universal" || n.tagNumber !== _T.INTEGER) throw ns.E(C, "InhibitAnyPolicy must be an INTEGER (RFC 5280 §4.2.1.14)");
+    if (n.tagClass !== "universal" || n.tagNumber !== _T.INTEGER) throw ns.E(C, "InhibitAnyPolicy must be an INTEGER (RFC 5280 sec. 4.2.1.14)");
     var v = readInt(n, C, "InhibitAnyPolicy");
     // Bounded before Number() narrowing (a skip count past the safe-integer
-    // range would round silently) — exact-or-rejected.
+    // range would round silently) -- exact-or-rejected.
     if (v < 0n || v > 2147483647n) throw ns.E(C, "InhibitAnyPolicy skip count must be a non-negative integer within range");
     return Number(v);
   }
 
-  // subjectAltName / issuerAltName ::= GeneralNames — decoded values surfaced.
+  // subjectAltName / issuerAltName ::= GeneralNames -- decoded values surfaced.
   function altName(buf) {
     var C = ns.prefix + "/bad-extension-value";
     var n = decodeTop(buf, C, "GeneralNames");
@@ -576,7 +746,7 @@ function certExtensionDecoders(ns) {
   function extKeyUsage(buf) {
     var C = ns.prefix + "/bad-extension-value";
     var kids = seqChildren(buf, C, "ExtKeyUsage");
-    if (kids.length < 1) throw ns.E(C, "ExtKeyUsage must contain at least one KeyPurposeId (RFC 5280 §4.2.1.12)");
+    if (kids.length < 1) throw ns.E(C, "ExtKeyUsage must contain at least one KeyPurposeId (RFC 5280 sec. 4.2.1.12)");
     return kids.map(function (k) {
       try { return asn1.read.oid(k); } catch (e) { throw ns.E(C, "ExtKeyUsage KeyPurposeId must be an OBJECT IDENTIFIER", e); }
     });
@@ -603,7 +773,7 @@ function certExtensionDecoders(ns) {
       else throw ns.E(C, "AuthorityKeyIdentifier has an unexpected field [" + k.tagNumber + "]");
     });
     if ((out.authorityCertIssuer === null) !== (out.authorityCertSerialNumber === null)) {
-      throw ns.E(C, "AuthorityKeyIdentifier authorityCertIssuer and authorityCertSerialNumber must both be present or both absent (RFC 5280 §4.2.1.1)");
+      throw ns.E(C, "AuthorityKeyIdentifier authorityCertIssuer and authorityCertSerialNumber must both be present or both absent (RFC 5280 sec. 4.2.1.1)");
     }
     return out;
   }
@@ -613,10 +783,10 @@ function certExtensionDecoders(ns) {
     var C = ns.prefix + "/bad-extension-value";
     var n = decodeTop(buf, C, "SubjectKeyIdentifier");
     try { return Buffer.concat([asn1.read.octetString(n)]); }
-    catch (e) { throw ns.E(C, "SubjectKeyIdentifier must be an OCTET STRING (RFC 5280 §4.2.1.2)", e); }
+    catch (e) { throw ns.E(C, "SubjectKeyIdentifier must be an OCTET STRING (RFC 5280 sec. 4.2.1.2)", e); }
   }
 
-  // signedCertificateTimestampList (RFC 6962 §3.3) — the extension value is a DER
+  // signedCertificateTimestampList (RFC 6962 sec. 3.3) -- the extension value is a DER
   // OCTET STRING wrapping the TLS-encoded SCT list. Delegated to the CT module,
   // which owns the inner peel + the bounded TLS decode; structure is decoded, the
   // signature stays raw and unverified. The registry contract is that a decoder
@@ -625,21 +795,29 @@ function certExtensionDecoders(ns) {
   function sctList(buf) {
     var C = ns.prefix + "/bad-extension-value";
     try { return ct.parseSctList(buf); }
-    catch (e) { throw ns.E(C, "malformed signedCertificateTimestampList extension value (RFC 6962 §3.3)", e); }
+    catch (e) { throw ns.E(C, "malformed signedCertificateTimestampList extension value (RFC 6962 sec. 3.3)", e); }
   }
 
-  // precertificatePoison (RFC 6962 §3.1) — a critical extension whose value is
+  // precertificatePoison (RFC 6962 sec. 3.1) -- a critical extension whose value is
   // exactly ASN.1 NULL (05 00), the marker distinguishing a precertificate from
   // a final certificate. Content-validated, not merely tag-checked.
   function precertPoison(buf) {
     var C = ns.prefix + "/bad-extension-value";
     var n = decodeTop(buf, C, "PrecertificatePoison");
     try { asn1.read.nullValue(n); }
-    catch (e) { throw ns.E(C, "the precertificate poison extension value must be ASN.1 NULL (RFC 6962 §3.1)", e); }
+    catch (e) { throw ns.E(C, "the precertificate poison extension value must be ASN.1 NULL (RFC 6962 sec. 3.1)", e); }
     return { poison: true };
   }
 
-  function O(nm) { return ns.oid.byName(nm); }
+  // byName returns undefined for an unregistered name; keying a decoder row
+  // under "undefined" would silently drop that extension from dispatch (the
+  // consumer would treat the extension as structurally unvalidated / absent),
+  // so a typo'd key-list entry must fail here, at module load.
+  function O(nm) {
+    var d = ns.oid.byName(nm);
+    if (typeof d !== "string") throw new TypeError("certExtensionDecoders: " + JSON.stringify(nm) + " is not a registered OID name");
+    return d;
+  }
   var byOid = {};
   byOid[O("basicConstraints")] = basicConstraints;
   byOid[O("keyUsage")] = keyUsage;
@@ -662,10 +840,10 @@ function certExtensionDecoders(ns) {
 // extnValue OCTET STRING }. `critical` is a universal BOOLEAN present-by-count
 // (not context-tagged), so the per-extension decode handles the 2-vs-3-child
 // shape directly; the seqOf centralizes the SEQUENCE / SIZE(1..MAX) assertion
-// and the RFC 5280 §4.2 per-OID uniqueness.
+// and the RFC 5280 sec. 4.2 per-OID uniqueness.
 function extension(ns) {
   return schema.decode(function (ext) {
-    // Extension ::= SEQUENCE { extnID, critical DEFAULT FALSE, extnValue } — a
+    // Extension ::= SEQUENCE { extnID, critical DEFAULT FALSE, extnValue } -- a
     // UNIVERSAL SEQUENCE of exactly 2 (critical omitted) or 3 children. A
     // context-tagged item (e.g. [5]{OID, OCTET STRING}) or a wrong child count
     // is malformed; assert the tag, don't just count children (fail closed).
@@ -677,7 +855,7 @@ function extension(ns) {
     var critical = false, valueNode;
     if (ext.children.length === 3) {
       critical = asn1.read.boolean(ext.children[1]);
-      // critical is BOOLEAN DEFAULT FALSE — DER omits the field when false, so an
+      // critical is BOOLEAN DEFAULT FALSE -- DER omits the field when false, so an
       // explicitly-encoded FALSE is a non-canonical form; reject it fail-closed.
       if (critical === false) throw ns.E(ns.prefix + "/bad-extension", "an explicit critical FALSE must be omitted (BOOLEAN DEFAULT FALSE)");
       valueNode = ext.children[2];
@@ -688,8 +866,8 @@ function extension(ns) {
   });
 }
 // opts.implicitTag (optional): read Extensions as a [tag] IMPLICIT SEQUENCE OF
-// Extension — the context tag REPLACES the universal SEQUENCE tag, for the CRMF
-// CertTemplate extensions [9] (RFC 4211 §5). With no opts the shape is a bare
+// Extension -- the context tag REPLACES the universal SEQUENCE tag, for the CRMF
+// CertTemplate extensions [9] (RFC 4211 sec. 5). With no opts the shape is a bare
 // universal SEQUENCE OF, byte-identical to every existing caller.
 function extensions(ns, opts) {
   opts = opts || {};
@@ -704,13 +882,13 @@ function extensions(ns, opts) {
 }
 
 // SubjectPublicKeyInfo ::= SEQUENCE { algorithm AlgorithmIdentifier,
-// subjectPublicKey BIT STRING } (RFC 5280 §4.1.2.7, RFC 2986 §4.1). Asserted as a
-// universal SEQUENCE — a context-tagged or SET-tagged constructed node carrying
+// subjectPublicKey BIT STRING } (RFC 5280 sec. 4.1.2.7, RFC 2986 sec. 4.1). Asserted as a
+// universal SEQUENCE -- a context-tagged or SET-tagged constructed node carrying
 // two well-formed children is NOT a SubjectPublicKeyInfo. Shared by the
 // certificate and CSR parsers.
 // opts.implicitTag (optional): read the SubjectPublicKeyInfo as a [tag] IMPLICIT
 // SEQUENCE (a context-class constructed node whose children are algorithm +
-// subjectPublicKey), for the CRMF CertTemplate publicKey [6] (RFC 4211 §5). With no
+// subjectPublicKey), for the CRMF CertTemplate publicKey [6] (RFC 4211 sec. 5). With no
 // opts the shape is a universal SEQUENCE, byte-identical to every existing caller.
 function spki(ns, opts) {
   opts = opts || {};
@@ -734,10 +912,10 @@ function spki(ns, opts) {
 }
 
 // Attribute ::= SEQUENCE { type OBJECT IDENTIFIER, values SET OF AttributeValue }.
-// AttributeValue is ANY, kept as raw DER (node.bytes) — an unrecognized attribute
+// AttributeValue is ANY, kept as raw DER (node.bytes) -- an unrecognized attribute
 // type never fails the parse. values is SET SIZE (1..MAX): an empty SET is
 // rejected; there is NO SET-OF uniqueness. Shared by the CSR (requested
-// attributes, RFC 2986 §4.1) and PKCS#8 (private-key attributes, RFC 5958 §2).
+// attributes, RFC 2986 sec. 4.1) and PKCS#8 (private-key attributes, RFC 5958 sec. 2).
 function attribute(ns) {
   return schema.seq([
     schema.field("type", schema.oidLeaf()),
@@ -772,7 +950,7 @@ function signedEnvelopeTbs(root) {
   return tbs;
 }
 
-// rootSequenceChildren(root, minLen, maxLen) — the shared root-shape guard the
+// rootSequenceChildren(root, minLen, maxLen) -- the shared root-shape guard the
 // format matches() detectors re-inline: returns root.children when root is a
 // NON-EMPTY universal SEQUENCE whose child count is within [minLen, maxLen]
 // (maxLen optional/Infinity), else null. Sibling of signedEnvelopeTbs.
@@ -789,20 +967,31 @@ function rootSequenceChildren(root, minLen, maxLen) {
 // bound parser so a format declares its configuration once and never re-writes the
 // coerce -> decode -> walk wrapper. `opts`: { pemLabel, PemError, ErrorClass,
 // prefix, what, topSchema, ns }.
+// The top-level schema's reject code -- the verdict for a DER that is not this
+// structure at all -- is `<prefix>/not-a-<structure>` (x509/not-a-certificate,
+// cms/not-a-content-info, ...). Three codes predate that convention and stay
+// frozen on the public error-code surface (ocsp/bad-ocsp-response,
+// tsp/bad-response, crmf/bad-cert-req-messages); a new format follows the
+// not-a-* form, never those precedents.
 function makeParser(opts) {
   return function (input) { return runParse(input, opts); };
 }
 
-// The X.509 SIGNED{ToBeSigned} macro (RFC 5280 §4.1.1.3): the outer
+// The X.509 SIGNED{ToBeSigned} macro (RFC 5280 sec. 4.1.1.3): the outer
 // SEQUENCE { toBeSigned, signatureAlgorithm AlgorithmIdentifier,
 // signatureValue BIT STRING } shared by Certificate, CertificateList and
 // CertificationRequest. `tbsSchema` parses the first element; the SEQUENCE-of-3
 // shape, the arity, the signature extraction and the raw tbs / outer-signature
 // bytes (for the cert/CRL outer==inner agreement check) are owned here once, and
 // each format's `opts.build(envelope, ctx)` shapes its own object from the
-// envelope. A CSR's build simply omits the agreement check — its CRI has no inner
-// signature AlgorithmIdentifier — so the omission is structural, not a copy that
+// envelope. A CSR's build simply omits the agreement check -- its CRI has no inner
+// signature AlgorithmIdentifier -- so the omission is structural, not a copy that
 // forgot a guard.
+// signatureValue is surfaced RAW as { unusedBits, bytes }: for the envelope
+// family, octet alignment is a verify-layer rule (path-validate rejects
+// unusedBits != 0 for certification paths; an external verifier must do the
+// same). A format with no in-tree verify layer (OCSP, CMP protection, CRMF
+// POP) instead enforces alignment at parse.
 function signedEnvelope(ns, tbsSchema, opts) {
   return schema.seq([
     schema.field("toBeSigned", tbsSchema),
@@ -833,11 +1022,13 @@ module.exports = {
   makeNS: makeNS,
   versionReader: versionReader,
   DN_SHORT: DN_SHORT,
+  time: time,
   algorithmIdentifier: algorithmIdentifier,
   spki: spki,
   makeParser: makeParser,
   signedEnvelopeTbs: signedEnvelopeTbs,
   rootSequenceChildren: rootSequenceChildren,
+  assertPolicyQualifiers: assertPolicyQualifiers,
   signedEnvelope: signedEnvelope,
   attrValueToString: attrValueToString,
   attributeTypeAndValue: attributeTypeAndValue,
@@ -845,6 +1036,10 @@ module.exports = {
   name: name,
   generalName: generalName,
   generalNames: generalNames,
+  generalizedTime: generalizedTime,
+  utf8Text: utf8Text,
+  rawNonEmptySequence: rawNonEmptySequence,
+  CRL_REASON_NAMES: CRL_REASON_NAMES,
   certExtensionDecoders: certExtensionDecoders,
   attribute: attribute,
   extension: extension,