relaton 3.0.1 → 3.0.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f81699e1df94f1c570f01c07b4114492b7cc5ae1e45715a4ced3ef51c1ba3fb6
-  data.tar.gz: 1ea38d455d16b512807a505bd558757ea0509a9392cf41c120f3fc10801beb87
+  metadata.gz: f15745a534625b27b3d013d7b3d6edd7e0185f1bcc1c545f7b44dc6059d9d113
+  data.tar.gz: ac15b5259abb3c980cd6bee9d10c5f2f0ae12fea955478ca28f7da82027ab6f6
 SHA512:
-  metadata.gz: 4d27294461f66031023053e1ec70a4c26b0c8fa068e6ed9d1224d1ab9c6cb366a2b2d8f85cd50305c4826733ab028817f70b96aebc267346d224b9be1fdf29be
-  data.tar.gz: a3c00ef5a93352f08a492624b531955494c8530a6b0d68f01dc748d58d8599304fa9ac743d157cfd2a6cc37b63664c781da75de5e199de70035ed133e70d0305
+  metadata.gz: de3f2c7182f6a2b1737a54e9168bcab856dbc81fe169f293b5b932f0224e8cb95187f506829230068ebc5308884e89a918515b03838f969c9c9870020442efff
+  data.tar.gz: a4bfe1d14667d8907eecd43fe880b02e7d15ceab20d2c6302c8f5c220d6c4398b3adefa056ff8815639686ad85674713c5c3a4759bd5ad57ffdd1e42b9bc39e1

data/lib/relaton/bib/converter/csl.rb

@@ -1,6 +1,7 @@
 # frozen_string_literal: true
 
-require "json"
+require "yeptris"
+
 
 module Relaton
   module Bib
@@ -43,7 +44,7 @@ module Relaton
             keywords = Array(@item.keyword).map { |k| k.content.to_s }.reject(&:empty?)
             out[:keyword] = keywords unless keywords.empty?
             out[:URL] = source_uri if source_uri
-            [out].to_json + "\n"
+            Yeptris::JSON.dump([out]) + "\n"
           end
 
           private

data/lib/relaton/bib/model/docidentifier.rb

@@ -44,6 +44,67 @@ module Relaton
       def to_all_parts!; end
 
       def remove_date!; end
+
+      # Docidentifier types with an SDO-specific :external and :internal form
+      # (relaton#49), mapped to the pubid flavor that renders both. The
+      # :external form carries the publisher ("IETF RFC 8341",
+      # "BSI PAS 2031:2019"); the :internal form is the one the SDO uses for its
+      # own documents, without it ("SP 800-113" for "NIST SP 800-113"). pubid
+      # decides per identifier, so "BS EN ISO 8848:2021" and "CGPM RES (1889)"
+      # are the same in both forms. Every other type answers its content for
+      # both forms: pubid would drop "ISO" from an ISO identifier.
+      #
+      # The table keys on `type`, not on a flavor subclass, so it also covers a
+      # relation's nested `Bib::ItemData` and the `Bib::Docidentifier` objects
+      # BIPM and IETF build (a subclass-typed collection would make lutaml's
+      # YAML serializer reject those).
+      FORM_PUBID = {
+        "NIST" => :Nist,
+        "BIPM" => :Bipm,
+        "IETF" => :Ietf,
+        "Internet-Draft" => :Ietf,
+        "BSI" => :Bsi,
+      }.freeze
+
+      FORMS = %i[external internal].freeze
+
+      # The parsed identifier. A pubid-backed flavor subclass overrides it.
+      def pubid; end
+
+      # @param form [Symbol, nil] nil for the stored content, :external for the
+      #   identifier with its publisher, :internal for the form the SDO uses
+      #   for its own documents
+      # @raise [ArgumentError] for an unknown form
+      def to_s(form: nil)
+        return content.to_s unless form
+
+        form = form.to_sym
+        unless FORMS.include?(form)
+          raise ArgumentError, "unknown identifier form: #{form.inspect}"
+        end
+
+        form_content(form).to_s
+      end
+
+      private
+
+      # pubid renders the form (`to_s(with_publisher:)`, pubid#620). A
+      # docidentifier with no parsed pubid (a plain `Bib::Docidentifier`) is
+      # parsed here; content that pubid cannot parse is returned unchanged.
+      def form_content(form)
+        flavor = FORM_PUBID[type]
+        id = flavor && (pubid || parse_form_pubid(flavor))
+        id ? id.to_s(with_publisher: form == :external) : content
+      end
+
+      def parse_form_pubid(flavor)
+        return if content.to_s.empty?
+
+        require "pubid"
+        ::Pubid.const_get(flavor).parse(content)
+      rescue StandardError
+        nil
+      end
     end
   end
 end

data/lib/relaton/bib.rb

@@ -5,6 +5,7 @@ require "bibtex"
 require "iso639"
 require "rfcxml"
 require "relaton/core"
+require "yeptris"
 require_relative "version"
 require_relative "bib/util"
 require_relative "bib/sanitizer"
@@ -41,7 +42,7 @@ module Relaton
     # @return [Hash{String=>String}] schema versions
     #
     def schema_versions
-      @@schema_versions ||= JSON.parse File.read(File.join(__dir__, "bib/versions.json"))
+      @@schema_versions ||= Yeptris::JSON.load File.read(File.join(__dir__, "bib/versions.json"))
     end
   end
 

data/lib/relaton/bipm/CLAUDE.md

@@ -144,6 +144,20 @@ demonstrated rather than argued.
 builds `index-v1` with them, and `id_parser_spec.rb` still covers them. Only
 `bibliography.rb` stopped requiring the file.
 
+### Internal identifier
+
+`docid.to_s(form: :internal)` drops the
+`BIPM` publisher (relaton#49): `BIPM SI Brochure Part 1` →
+`SI Brochure Part 1`. Committee ids (`CGPM RES (1889)`, `CCTF REC 2 (2009)`)
+have no publisher token and are the same in both forms. The rule is the `"BIPM"` row of
+`Bib::Docidentifier::FORM_PUBID`: BIPM docidentifiers carry no pubid,
+so the content is parsed with `Pubid::Bipm` at call time and rendered with
+`to_s(with_publisher: false)`; unparseable content is returned unchanged. It
+also holds for a relation's nested item. Do not move it into a `Bipm::Docidentifier`
+subclass: typing the item's `docidentifier` collection with a subclass makes
+YAML serialization reject the `Bib::Docidentifier` objects the parsers build
+(`IncorrectModelError`).
+
 ### Data Sources
 
 The gem fetches from three external datasets:

data/lib/relaton/bsi/model/docidentifier.rb

@@ -44,10 +44,6 @@ module Relaton
         c if c.is_a?(::Pubid::Bsi::Identifier) || c.is_a?(::Pubid::AllParts)
       end
 
-      def to_s
-        content.to_s
-      end
-
       # Strip the publication date to build a most-recent (undated) reference.
       # pubid's `exclude` propagates into nested identifiers (so a consolidated
       # base date is dropped while the amendment is kept); `:month` is excluded

data/lib/relaton/calconnect/data_fetcher.rb

@@ -1,6 +1,6 @@
+require "yeptris"
 # frozen_string_literal:true
 
-require "json"
 require "mechanize"
 require "relaton/core"
 require "relaton/index"
@@ -54,7 +54,7 @@ module Relaton::Calconnect
       # 304 Not Modified — nothing changed since the last fetch
       return if resp.code == "304"
 
-      data = JSON.parse resp.body
+      data = Relaton::Json.parse_external resp.body
       all_success = true
       Array(data["documents"]).each { |doc| all_success &&= parse_page doc }
       self.etag = resp.response["etag"] if all_success

data/lib/relaton/ccsds/data/fetcher.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 require "json"
 require "mechanize"
 require "relaton/index"
@@ -43,7 +44,7 @@ module Relaton
       #
       def fetch_docs(url)
         resp = agent.get(url)
-        json = JSON.parse resp.body.match(/const config = (.*);/)[1]
+        json = Relaton::Json.parse_external resp.body.match(/const config = (.*);/)[1]
         @array = json["data"].map { |doc| parse_and_save doc, json["data"] }
       end
 

data/lib/relaton/ccsds/model/docidentifier.rb

@@ -60,10 +60,6 @@ module Relaton
         original_content
       end
 
-      def to_s
-        content.to_s
-      end
-
       def to_all_parts!
         return if !@pubid || @pubid.all_parts?
 

data/lib/relaton/core/data_fetcher.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 require "fileutils"
 
 module Relaton
@@ -361,7 +362,7 @@ module Relaton
       def entry_digest(entry)
         require "digest"
         require "json"
-        Digest::SHA256.hexdigest JSON.generate(entry)
+        Digest::SHA256.hexdigest Yeptris::JSON.dump(entry)
       end
 
       # Write, clobbering whatever is there. The caller has already established

data/lib/relaton/core.rb

@@ -1,4 +1,11 @@
 require "relaton/logger"
+
+module Relaton
+  # Registered here as well as in the root entry: flavor files and their
+  # specs load directly via require "relaton/<flavor>", never through
+  # relaton.rb, and the boundary parsers need this constant either way.
+  autoload :Json, "relaton/json"
+end
 require_relative "version"
 require_relative "core/request_error"
 require_relative "core/array_wrapper"

data/lib/relaton/db/cache.rb

@@ -1,6 +1,6 @@
+require "yeptris"
 require "fileutils"
 require "digest"
-require "json"
 require "date"
 require "lutaml/store"
 require_relative "cache_entry"
@@ -520,7 +520,7 @@ module Relaton
       def canonical(key)
         return key if key.is_a? String
 
-        JSON.parse JSON.generate(key.to_hash)
+        Yeptris::JSON.load Yeptris::JSON.dump(key.to_hash)
       end
 
       def row_key(row)
@@ -556,7 +556,7 @@ module Relaton
 
       # A digest of the canonical key, independent of the hash order.
       def digest(key)
-        Digest::SHA256.hexdigest JSON.generate(canonical_sorted(key))
+        Digest::SHA256.hexdigest Yeptris::JSON.dump(canonical_sorted(key))
       end
 
       def canonical_sorted(key)

data/lib/relaton/db.rb

@@ -127,7 +127,15 @@ module Relaton
           rescue Relaton::RequestError => e
             args[3].call e
           rescue StandardError => e
-            Util.error "`#{args[0]}` -- #{e.message}"
+            # An unparseable reference ("ISO Standards Tag Set" in a
+            # vocabulary corpus) is a not-found input, not a failure: the
+            # citation keeps its literal text and the build continues
+            # (relaton#412). Real errors keep the ERROR level.
+            if unreadable_reference?(e)
+              Util.info "`#{args[0]}` -- #{e.message}"
+            else
+              Util.error "`#{args[0]}` -- #{e.message}"
+            end
             args[3].call nil
           end
           @queues[stdclass] =
@@ -213,6 +221,14 @@ module Relaton
 
     private
 
+    # True when the error is a flavor's grammar refusing to read the
+    # reference. Both are raised by the flavors' pubid/parslet parses; a
+    # reference they cannot read is not-found, not a failure (relaton#412).
+    def unreadable_reference?(error)
+      (defined?(Pubid::Errors::Error) && error.is_a?(Pubid::Errors::Error)) ||
+        (defined?(Parslet::ParseFailed) && error.is_a?(Parslet::ParseFailed))
+    end
+
     # The flavor routing picks for `fetch_async`'s per-flavor queue; nil
     # (logged) when no flavor recognizes the reference.
     def async_route(ref)
@@ -309,6 +325,9 @@ module Relaton
           )
         end
       end
+      # Nothing resolved: answer not-found, not a stub that holds only the
+      # reference (relaton-un#14, metanorma-ietf#205).
+      doc.relation.empty? ? nil : doc
     end
 
     # The legacy string cache key, for a processor with no pubid class. The

data/lib/relaton/doi/crossref.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 require "mechanize"
 
 module Relaton
@@ -58,7 +59,7 @@ module Relaton
         url = "https://api.crossref.org/works/#{CGI.escape(id)}"
         loop do
           resp = agent.get url
-          work = JSON.parse resp.body
+          work = Relaton::Json.parse_external resp.body
           return work["message"] if work["status"] == "ok"
 
           if n > 1

data/lib/relaton/doi/parser.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 module Relaton
   module Doi
     class Parser
@@ -856,7 +857,7 @@ module Relaton
         throttled = 0
         begin
           resp = Crossref.agent.get url
-          JSON.parse(resp.body).dig("message", "items")
+          Relaton::Json.parse_external(resp.body).dig("message", "items")
         rescue Mechanize::ResponseCodeError => e
           # 429 must be handled before the 4xx nil return: a throttle is not
           # "no such record". Returning nil here made a rate-limited lookup look
@@ -881,7 +882,7 @@ module Relaton
           retries += 1
           retry if retries <= MAX_RETRIES
           raise Relaton::RequestError, "Crossref network error after #{MAX_RETRIES} retries: #{e.message}"
-        rescue JSON::ParserError => e
+        rescue ::Yeptris::JSON::ParseError => e
           raise Relaton::RequestError, "Crossref JSON parsing error: #{e.message}"
         end
       end

data/lib/relaton/etsi/data_fetcher.rb

@@ -1,5 +1,5 @@
+require "yeptris"
 require "date"
-require "json"
 require "mechanize"
 require "relaton/core"
 require "relaton/index"
@@ -146,11 +146,11 @@ module Relaton
       # @raise [BadPage] when the body is not a JSON array
       #
       def parse_page(page, body)
-        records = JSON.parse(body)
+        records = Relaton::Json.parse_external(body)
         return records if records.is_a?(Array)
 
         raise BadPage, bad_page_message(page, body)
-      rescue JSON::ParserError
+      rescue ::Yeptris::JSON::ParseError
         raise BadPage, bad_page_message(page, body)
       end
 

data/lib/relaton/gb/sec_scraper.rb

@@ -1,8 +1,8 @@
 # encoding: UTF-8
 # frozen_string_literal: true
 
+
 require "net/http"
-require "json"
 require "moxml"
 require_relative "scraper"
 require_relative "item"
@@ -24,8 +24,8 @@ module Relaton
           uri = URI "https://hbba.sacinfo.org.cn/stdQueryList"
           resp = Net::HTTP.post uri, URI.encode_www_form({ key: text }),
                                 "Content-Type" => "application/x-www-form-urlencoded"
-          # res = JSON.parse Net::HTTP.get(uri)
-          json = JSON.parse resp.body
+          # res = Relaton::Json.parse_external Net::HTTP.get(uri)
+          json = Relaton::Json.parse_external resp.body
           hits = json["records"].map do |h|
             Hit.new pid: h["pk"], docref: h["code"], status: h["status"], scraper: self
           end

data/lib/relaton/iec/data_fetcher.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 require "relaton/core"
 require_relative "../iec"
 require_relative "data_parser"
@@ -135,7 +136,7 @@ module Relaton
             Util.warn "#{res.body}"
             break
           end
-          json = JSON.parse res.body
+          json = Relaton::Json.parse_external res.body
           json[list_key].each { |pub| fetch_pub pub }
           page += 1
           next_page = res["link"]&.include? "rel=\"last\""
@@ -191,7 +192,7 @@ module Relaton
           res = Net::HTTP.start(uri.hostname, uri.port, use_ssl: true) do |http|
             http.request req
           end
-          JSON.parse(res.body)["access_token"]
+          Relaton::Json.parse_external(res.body)["access_token"]
         end
       end
 

data/lib/relaton/iec/model/docidentifier.rb

@@ -46,10 +46,6 @@ module Relaton
         original_content
       end
 
-      def to_s
-        content.to_s
-      end
-
       # `remove_part!`/`remove_date!` are unnecessary here: `to_all_parts`'s
       # own identity computation already excludes part/date (down the whole
       # base chain, so a supplement's own year is dropped too — a disclosed

data/lib/relaton/ietf/CLAUDE.md

@@ -36,7 +36,7 @@ Converter classes use `Bib::NamespaceHelper` which resolves `namespace` by takin
 
 ### Key Flows
 
-1. **Single document lookup**: `Processor#get` → `Bibliography.get(code)` → `Scraper.scrape_page` → parses the reference with `Pubid::Ietf` → searches the one combined `index-v2` from `relaton-data-ietf` → fetches that row's YAML
+1. **Single document lookup**: `Processor#get` → `Bibliography.get(code)` → `Scraper.scrape_page` → parses the reference with `Pubid::Ietf` → searches the one combined `index-v2` from `relaton-data-ietf` → fetches that row's YAML; a versioned draft the index lacks is read from bib.ietf.org (see below)
 
 2. **Bulk data fetching**: `DataFetcher` extends `Relaton::Core::DataFetcher` with three datasets:
    - `ietf-rfcsubseries` / `ietf-rfc-entries`: parse `rfc-index.xml` via `Rfc::Index` / `Rfc::Entry#to_item`
@@ -119,6 +119,30 @@ The old plain-string index matched that by substring; matching is exact now.
 A reference pubid cannot parse returns nil, so `Bibliography.get "CN 8341"` logs
 "Not found." rather than raising.
 
+### An index miss on a versioned draft reads bib.ietf.org (relaton-ietf#125)
+
+The data repo crawls the rsync `bibxml-ids` snapshot, and that snapshot keeps
+only some versions of each draft: `draft-ietf-vcarddav-vcardrev` has only `-22`,
+so `-06` was "Not found." When `fetch_doc` misses and the query is a
+`Pubid::Ietf::Identifiers::InternetDraft` **with a version**, `fetch_bibxml` reads
+`https://bib.ietf.org/public/rfc/bibxml-ids/reference.I-D.<draft>-NN.xml`
+(`BIBXML_IDS`). That service resolves any version the datatracker knows and
+answers 404 otherwise; the datatracker's own `doc/bibxml3/...` URL answers 404
+for these old versions. The parse is the crawler's: `BibXMLParser.to_utf8`
+(the CP1252 repair, shared with `DataFetcher#read_bibxml`), `BibXMLParser.parse`,
+and the version from the identifier, as `process_singleton` sets it.
+
+Only a 404 is "not found". Any other non-200 (429, 5xx) raises
+`Relaton::RequestError`, so `Db#net_retry` retries it and `Db::Cache` does not
+store an outage as a not-found row. A 200 whose body does not parse logs a
+warning and returns nil; that rescue names the parse errors and is not
+`StandardError`, so a transport error still becomes `RequestError`.
+
+RFCs, sub-series and unversioned drafts do **not** fall back: the index holds
+all of them, so a miss is a real "not found". A record read this way has no
+`updates`/`updatedBy` neighbours — the crawl links those, and one fetch cannot.
+Old bibxml carries no `target`, so such a record has a `TXT` URI and no `src`.
+
 ### The producer: `DataFetcher` writes `index-v2`
 
 `DataFetcher` builds the index during the crawl, from the identifier it already
@@ -294,7 +318,8 @@ smart quotes and accented Latin letters. `File.read(encoding: "UTF-8")` only
 *tags* a string; it validates nothing, so those reached lutaml as invalid UTF-8
 and `Rfcxml::V3::Reference.from_xml` raised `Lutaml::Model::InvalidFormatError`.
 
-`read_bibxml` transcodes them (`data_fetcher.rb`). Three things make that the
+`read_bibxml` transcodes them through `BibXMLParser.to_utf8`
+(`bibxml_parser.rb`, shared with the scraper's bib.ietf.org fallback). Three things make that the
 right call rather than `scrub`:
 
 - **All of them decode losslessly as CP1252** — 125 of the ~122k published files

data/lib/relaton/ietf/bibxml_parser.rb

@@ -13,6 +13,44 @@ module Relaton
         FromRfc.new(rfc).transform
       end
 
+      #
+      # Decode bibxml bytes as UTF-8, recovering Windows-1252 bytes.
+      #
+      # These files declare `encoding='UTF-8'` but some carry CP1252 — smart
+      # quotes and accented Latin letters. `File.read(encoding: "UTF-8")` only
+      # *tags* the string, so those reach lutaml as invalid UTF-8 and it raises.
+      #
+      # `scrub` with a block transcodes each invalid *run* as CP1252 while
+      # leaving valid UTF-8 untouched. Both halves matter:
+      #
+      # * Not plain `scrub`, which substitutes U+FFFD: `client’s` would become
+      #   `client�s` and `Muñoz` `Mu�oz` — author surnames included.
+      #   The damage is length-preserving, so a length check will not catch it.
+      # * Not a whole-file CP1252 re-decode, which mangles a file that is
+      #   genuinely UTF-8 apart from one stray byte: `Muñoz café ’` would come
+      #   back as `Muñoz café ’`, silently, since the result is valid UTF-8 and
+      #   nothing raises. No such file exists in today's corpus (0 of the 125
+      #   affected contain valid multi-byte UTF-8) but it grows daily, and
+      #   "decodes losslessly as CP1252" is weak evidence of correctness —
+      #   CP1252 maps 251 of 256 byte values.
+      #
+      # `undef: :replace` covers the five bytes CP1252 leaves undefined
+      # (0x81 0x8D 0x8F 0x90 0x9D), so this returns valid UTF-8 rather than
+      # raising and costing the whole file.
+      #
+      # @param bytes [String] raw bytes, any encoding tag
+      # @return [String] UTF-8, valid encoding
+      #
+      def to_utf8(bytes)
+        utf8 = bytes.dup.force_encoding(Encoding::UTF_8)
+        return utf8 if utf8.valid_encoding?
+
+        utf8.scrub do |bad|
+          bad.force_encoding(Encoding::WINDOWS_1252)
+             .encode(Encoding::UTF_8, undef: :replace)
+        end
+      end
+
       def pubid_type(id)
         pref = id.match(/^(\S+)/)[1]
         case pref

data/lib/relaton/ietf/data_fetcher.rb

@@ -216,41 +216,14 @@ module Relaton
       end
 
       #
-      # Read a bibxml file as UTF-8, recovering Windows-1252 bytes.
-      #
-      # These files declare `encoding='UTF-8'` but some carry CP1252 — smart
-      # quotes and accented Latin letters. `File.read(encoding: "UTF-8")` only
-      # *tags* the string, so those reach lutaml as invalid UTF-8 and it raises.
-      #
-      # `scrub` with a block transcodes each invalid *run* as CP1252 while
-      # leaving valid UTF-8 untouched. Both halves matter:
-      #
-      # * Not plain `scrub`, which substitutes U+FFFD: `client’s` would become
-      #   `client\uFFFDs` and `Muñoz` `Mu\uFFFDoz` — author surnames included.
-      #   The damage is length-preserving, so a length check will not catch it.
-      # * Not a whole-file CP1252 re-decode, which mangles a file that is
-      #   genuinely UTF-8 apart from one stray byte: `Muñoz café ’` would come
-      #   back as `Muñoz café ’`, silently, since the result is valid UTF-8 and
-      #   nothing raises. No such file exists in today's corpus (0 of the 125
-      #   affected contain valid multi-byte UTF-8) but it grows daily, and
-      #   "decodes losslessly as CP1252" is weak evidence of correctness —
-      #   CP1252 maps 251 of 256 byte values.
-      #
-      # `undef: :replace` covers the five bytes CP1252 leaves undefined
-      # (0x81 0x8D 0x8F 0x90 0x9D), so this returns valid UTF-8 rather than
-      # raising and costing the whole file.
+      # Read a bibxml file as UTF-8, recovering Windows-1252 bytes. The repair
+      # is `BibXMLParser.to_utf8`, shared with `Scraper`'s bib.ietf.org fallback.
       #
       # @param path [String]
       # @return [String] UTF-8, valid encoding
       #
       def read_bibxml(path)
-        utf8 = File.binread(path).force_encoding(Encoding::UTF_8)
-        return utf8 if utf8.valid_encoding?
-
-        utf8.scrub do |bad|
-          bad.force_encoding(Encoding::WINDOWS_1252)
-             .encode(Encoding::UTF_8, undef: :replace)
-        end
+        BibXMLParser.to_utf8 File.binread(path)
       end
 
       #

data/lib/relaton/ietf/scraper.rb

@@ -2,6 +2,7 @@
 
 require "pubid"
 require "pubid/ietf"
+require_relative "bibxml_parser"
 
 module Relaton
   module Ietf
@@ -15,6 +16,13 @@ module Relaton
       # publishing their `index-v1` for released relatons, untouched.
       IETF = "https://raw.githubusercontent.com/relaton/relaton-data-ietf/main/"
 
+      # The IETF bibxml service, read when the index has no row for a versioned
+      # Internet-Draft (relaton-ietf#125). The data repo crawls the rsync
+      # `bibxml-ids` snapshot, which keeps only some versions of each draft
+      # (`draft-ietf-vcarddav-vcardrev`: only -22); this service resolves any
+      # version the datatracker knows, and answers 404 for one it does not.
+      BIBXML_IDS = "https://bib.ietf.org/public/rfc/bibxml-ids/"
+
       # @param ref [String, Pubid::Ietf::Identifier] a reference, or its parse
       #   from Relaton::Db (relaton#205), which is used as it is
       # @return [Relaton::Ietf::ItemData, nil]
@@ -24,8 +32,9 @@ module Relaton
 
         fetch_doc id
       rescue Timeout::Error, Errno::EINVAL, Errno::ECONNRESET, EOFError,
-             Net::HTTPBadResponse, Net::HTTPHeaderSyntaxError,
-             Net::ProtocolError, SocketError
+             Errno::ECONNREFUSED, Errno::EHOSTUNREACH, Errno::ENETUNREACH,
+             OpenSSL::SSL::SSLError, Net::HTTPBadResponse,
+             Net::HTTPHeaderSyntaxError, Net::ProtocolError, SocketError
         raise Relaton::RequestError, "No document found for #{ref} reference"
       end
 
@@ -44,7 +53,55 @@ module Relaton
       # such pairs in the first 40k rows of the index fixture).
       def fetch_doc(id)
         row = index.search(id, exact: true).first
-        get_page "#{IETF}#{row[:file]}" if row
+        return get_page "#{IETF}#{row[:file]}" if row
+
+        fetch_bibxml id if versioned_draft? id
+      end
+
+      # Only a versioned draft falls back. The index holds every RFC, sub-series
+      # and unversioned draft aggregator the data repo has, so a miss on one of
+      # those is a real "not found", and bib.ietf.org has no aggregator to offer.
+      def versioned_draft?(id)
+        id.is_a?(::Pubid::Ietf::Identifiers::InternetDraft) && !id.version.to_s.empty?
+      end
+
+      #
+      # Read one draft version from bib.ietf.org and parse it the way
+      # `DataFetcher#process_singleton` parses a crawled file: the same
+      # CP1252 repair, the same parser, and the version from the identifier.
+      #
+      # A 404 is the service's "no such version". Any other non-200 (a 429, a
+      # 5xx) is push-back or an outage, so it raises `Relaton::RequestError`:
+      # `Db#net_retry` then retries it, and the Db cache does not store it as a
+      # not-found.
+      #
+      # @param id [Pubid::Ietf::Identifiers::InternetDraft]
+      # @return [Relaton::Ietf::ItemData, nil] nil on a 404 or an unreadable body
+      #
+      def fetch_bibxml(id)
+        res = Net::HTTP.get_response URI("#{BIBXML_IDS}reference.I-D.#{id}.xml")
+        return if res.code == "404"
+        raise Relaton::RequestError, "#{BIBXML_IDS} answered #{res.code} for #{id}" unless res.code == "200"
+
+        bib = parse_bibxml(res.body, id)
+        return unless bib
+
+        bib.version = [Bib::Version.new(draft: id.version.to_s)]
+        bib.fetched = Date.today.to_s
+        bib
+      end
+
+      # A 200 that is not bibxml (an HTML error page, a truncated body) costs
+      # this one document, not the caller's compile. These are the errors the
+      # parse raises on such a body: lutaml's `InvalidFormatError`, and the
+      # converter's `NoMethodError` on a document with no `<front>` and
+      # `ArgumentError` from its regexes. They are named, not `StandardError`,
+      # so a transport error still reaches `scrape_page`'s rescue.
+      def parse_bibxml(body, id)
+        BibXMLParser.parse BibXMLParser.to_utf8(body)
+      rescue Lutaml::Model::InvalidFormatError, NoMethodError, ArgumentError => e
+        Util.warn "Unreadable bibxml for `#{id}`: #{e.message.to_s.lines.first.to_s.strip}"
+        nil
       end
 
       #

data/lib/relaton/ietf/wg_name_resolver.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-require "json"
+
 require "net/http"
 
 module Relaton
@@ -20,7 +20,7 @@ module Relaton
           response = Net::HTTP.get_response(uri)
           break unless response.is_a?(Net::HTTPSuccess)
 
-          data = JSON.parse(response.body)
+          data = Relaton::Json.parse_external(response.body)
           objects = data["objects"] || []
           break if objects.empty?
 

data/lib/relaton/iho/CLAUDE.md

@@ -55,6 +55,10 @@ In `item.rb`, `require_relative "relation"` is placed **after** the `Item` class
 
 `Bibliography.search`/`.get` fetches YAML from the relaton-data-iho GitHub repo via `Relaton::Index`, then deserializes into model objects.
 
+**Editions.** The pubid `version` is the edition (`IHO B-12 3.0.0`; the grammar takes `a.b.c` only). A reference with no version answers with the **latest** edition: `version_key` compares the segments as integers (`2.10.0` after `2.9.0`). It used to be `min_by` on the string, which returned the oldest.
+
+**The edition reaches the index only through the committed row.** A record prints `B-12` and keeps its edition in `edition:`, so `relaton index` cannot read a version from the docid. `MachineIndex#refine` (relaton-cli) takes the committed `index-v3.yaml` row when it adds only a `version` to the parse. Before that, the Pages index (generated 2026-10-29) had no `version` in any of its 325 rows, so every versioned reference answered not-found (relaton-data-iho#20). The suite did not see it: `spec/fixtures/index-v3.zip` is the committed index, which has versions.
+
 ### Test Infrastructure
 
 - **VCR cassettes** (`spec/vcr_cassettes/`): recorded HTTP responses for offline testing

data/lib/relaton/iho/bibliography.rb

@@ -21,7 +21,7 @@ module Relaton
           # Pubid::Iho::Identifiers::* (Relaton::Index deserialized it via the
           # `pubid_class` passed in `#index`), so `pubid_match?` compares pubids.
           row = index.search(pubid) { |r| pubid_match?(r[:id], pubid) }
-                     .min_by { |r| r[:id].version.to_s }
+                     .max_by { |r| version_key(r[:id]) }
           unless row
             Util.info "Not found.", key: pubid.to_s
             return
@@ -96,6 +96,12 @@ module Relaton
           )
         end
 
+        # The sort key for the latest edition: segments compare as integers,
+        # so 2.10.0 is later than 2.9.0. A row with no version is the oldest.
+        def version_key(id)
+          id.version.to_s.split(".").map(&:to_i)
+        end
+
         # Both `row_id` and `query` are Pubid::Iho::Identifiers::* instances.
         # Class identity pins the series (S/P/M/B/C). Subdivision keys
         # (part/appendix/annex/supplement) use strict equality — a nil query

data/lib/relaton/index/dataset_lock.rb

@@ -1,7 +1,8 @@
 # frozen_string_literal: true
 
+require "yeptris"
+
 require "digest"
-require "json"
 require "net/http"
 require "time"
 require "uri"
@@ -91,8 +92,8 @@ module Relaton
         end
         raise Error, "cannot fetch #{uri}: HTTP #{res.code}" unless res.code == "200"
 
-        JSON.parse(res.body)
-      rescue JSON::ParserError => e
+        Yeptris::JSON.load(res.body)
+      rescue Yeptris::JSON::ParseError => e
         raise Error, "malformed manifest at #{uri}: #{e.message}"
       end
 

data/lib/relaton/index/shard_source.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 require "json"
 require "net/http"
 require "openssl"
@@ -174,8 +175,8 @@ module Relaton
       end
 
       def parse_json(body, path)
-        JSON.parse(body)
-      rescue JSON::ParserError => e
+        Yeptris::JSON.load(body)
+      rescue Yeptris::JSON::ParseError => e
         raise Error, "cannot parse `#{url(path)}`: #{e.message}"
       end
 

data/lib/relaton/index/sqlite_backend.rb

@@ -1,5 +1,7 @@
 # frozen_string_literal: true
 
+require "yeptris"
+
 require "sqlite3"
 
 module Relaton
@@ -34,7 +36,7 @@ module Relaton
         @db.transaction do
           @db.execute("DELETE FROM index_rows")
           @db.prepare("INSERT INTO index_rows (number, sort_key, id_json, file) VALUES (?, ?, ?, ?)") do |stmt|
-            rows_enum.each { |(number, sort_key, id_hash, file)| stmt.execute(number, sort_key, JSON.generate(id_hash), file) }
+            rows_enum.each { |(number, sort_key, id_hash, file)| stmt.execute(number, sort_key, Yeptris::JSON.dump(id_hash), file) }
           end
           meta_set("schema_version", SCHEMA_VERSION)
         end
@@ -44,14 +46,14 @@ module Relaton
       def bucket(number)
         rows = []
         @db.execute("SELECT id_json, file FROM index_rows WHERE number = ? ORDER BY rowid", [number]) do |row|
-          rows << { id: JSON.parse(row[0]), file: row[1] }
+          rows << { id: Yeptris::JSON.load(row[0]), file: row[1] }
         end
         rows
       end
 
       def each_row
         @db.execute("SELECT id_json, file FROM index_rows ORDER BY rowid") do |row|
-          yield({ id: JSON.parse(row[0]), file: row[1] })
+          yield({ id: Yeptris::JSON.load(row[0]), file: row[1] })
         end
       end
 

data/lib/relaton/isbn/open_library.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 module Relaton
   module Isbn
     #
@@ -37,7 +38,7 @@ module Relaton
         response = Net::HTTP.get_response uri
         return unless response.is_a? Net::HTTPSuccess
 
-        data = JSON.parse response.body
+        data = Relaton::Json.parse_external response.body
         return unless data["records"]&.any?
 
         data["records"].first.last

data/lib/relaton/iso/data_fetcher.rb

@@ -1,5 +1,5 @@
+require "yeptris"
 require "fileutils"
-require "json"
 require "net/http"
 require "tmpdir"
 require_relative "../iso"
@@ -168,7 +168,7 @@ module Relaton
         amend_map = Hash.new { |h, k| h[k] = [] }
         date_map = {}
         File.foreach(path, encoding: "UTF-8") do |line|
-          rec = JSON.parse(line)
+          rec = Relaton::Json.parse_external(line)
           id = rec["id"]
           ref = normalize_reference(rec["reference"])
           next unless ref
@@ -179,7 +179,7 @@ module Relaton
           if rec["supplementType"] && (base = amend_base(ref))
             amend_map[base] << ref
           end
-        rescue JSON::ParserError
+        rescue ::JSON::ParserError
           next
         end
         Util.info "Indexed #{ref_map.size} references; " \
@@ -216,11 +216,11 @@ module Relaton
         path = download_tc_dataset
         map = {}
         File.foreach(path, encoding: "UTF-8") do |line|
-          rec = JSON.parse(line)
+          rec = Relaton::Json.parse_external(line)
           ref = rec["reference"]
           title = rec["title"]
           map[ref] = title if ref && title.is_a?(Hash)
-        rescue JSON::ParserError
+        rescue ::JSON::ParserError
           next
         end
         Util.info "Indexed #{map.size} committees."
@@ -231,7 +231,7 @@ module Relaton
         Util.info "Parsing records..."
         count = 0
         File.foreach(path, encoding: "UTF-8") do |line|
-          rec = JSON.parse(line)
+          rec = Relaton::Json.parse_external(line)
           next unless ingestable?(rec["reference"])
 
           fetch_pub(rec, ref_index, tc_index, amend_index, date_index)

data/lib/relaton/iso/model/docidentifier.rb

@@ -64,10 +64,6 @@ module Relaton
         original_content
       end
 
-      def to_s
-        content.to_s
-      end
-
       # `remove_part!`/`remove_date!` are unnecessary here: `to_all_parts`'s
       # own identity computation already excludes part/date (down the whole
       # base chain, so a supplement's own year is dropped too — a disclosed

data/lib/relaton/itu/data_fetcher.rb

@@ -1,5 +1,5 @@
+require "yeptris"
 require "net/http"
-require "json"
 require "uri"
 require "mechanize"
 require_relative "../itu"
@@ -542,7 +542,7 @@ module Relaton
         request["User-Agent"] = USER_AGENT
 
         response = http.request(request)
-        json = JSON.parse(response.body)
+        json = Relaton::Json.parse_external(response.body)
         json["Data"] || []
       end
     end

data/lib/relaton/itu/hit_collection.rb

@@ -1,7 +1,7 @@
 # frozen_string_literal: true
 
+
 require "cgi"
-require "json"
 require_relative "hit"
 
 module Relaton
@@ -239,8 +239,8 @@ module Relaton
       # @return [Array<Hash>] editions of the recommendation
       def editions(idrec)
         # a `null` body parses to nil, so `|| []` is not redundant
-        JSON.parse(agent.get(format(RECEDITIONS_URL, idrec: idrec)).body) || []
-      rescue JSON::ParserError
+        Relaton::Json.parse_external(agent.get(format(RECEDITIONS_URL, idrec: idrec)).body) || []
+      rescue ::JSON::ParserError
         []
       end
 

data/lib/relaton/itu/recommendation_fields.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-require "json"
+
 require_relative "family_cache"
 
 module Relaton
@@ -251,7 +251,7 @@ module Relaton
       end
 
       def get_data(url)
-        JSON.parse request_document(url).body
+        Relaton::Json.parse_external request_document(url).body
       end
 
       def request_document(url)

data/lib/relaton/jis/data_fetcher.rb

@@ -1,5 +1,6 @@
 # frozen_string_literal: true
 
+
 require_relative "../jis"
 require_relative "scraper"
 
@@ -92,7 +93,7 @@ module Relaton
         body = { record: 0, dantai: "JIS", searchtype2: 1,
                  status_1: 1, status_2: 2 } # rubocop:disable Naming/VariableNumber
         resp = agent.post "#{URL}W11M0270/index", body
-        disp = JSON.parse resp.body
+        disp = Relaton::Json.parse_external resp.body
         @initial_time = Time.now
         disp["status"] || Util.warn("No results found for JIS")
       end

data/lib/relaton/json.rb

@@ -0,0 +1,21 @@
+require "yeptris"
+
+module Relaton
+  # JSON at the system boundary — data we do not control. Everything we
+  # generate or own (index manifests, shards, the db cache, request bodies,
+  # the schema-version table) parses with Yeptris::JSON's default strict
+  # surface.
+  module Json
+    module_function
+
+    # Third-party feeds carry duplicate keys in the wild (Crossref book
+    # records keyed by ISBN, NIST's scraped pubs export keyed by "&nbsp",
+    # the UN search API's blank keys, CCSDS's embedded config); parsing them
+    # the historical way is last-wins. yeptris 0.6.51 carries the boundary
+    # opt-out (yeptris-ruby#335/#339), so the whole stack stays on yeptris —
+    # no stdlib fallback anywhere.
+    def parse_external(source)
+      ::Yeptris::JSON.load(source, allow_duplicate_key: true)
+    end
+  end
+end

data/lib/relaton/logger/channels/gh_issue.rb

@@ -1,5 +1,5 @@
+require "yeptris"
 require "net/http"
-require "json"
 
 module Relaton
   module Logger
@@ -70,7 +70,7 @@ module Relaton
           responce = get(uri)
           return nil unless responce.code.to_i == 200
 
-          items = JSON.parse(responce.body)["items"] || []
+          items = Yeptris::JSON.load(responce.body)["items"] || []
           items.find { |i| i["title"] == @title }&.fetch("number", nil)
         rescue StandardError => e
           puts "Failed to search for an open `#{@title}` issue: #{e.message}"
@@ -88,7 +88,7 @@ module Relaton
           http.use_ssl = true
 
           request = Net::HTTP::Post.new(uri.request_uri, headers)
-          request.body = body.to_json
+          request.body = Yeptris::JSON.dump(body)
 
           http.request(request)
         end

data/lib/relaton/logger/formatter_json.rb

@@ -1,8 +1,9 @@
+require "yeptris"
 module Relaton::Logger
   class FormatterJSON < ::Logger::Formatter
     def call(severity, _datetime, progname, msg, **args)
       hash = { prog: progname, message: msg, severity: severity }.merge(args)
-      "#{hash.to_json}\n"
+      Yeptris::JSON.dump(hash) + "\n"
     end
   end
 end

data/lib/relaton/nist/CLAUDE.md

@@ -64,6 +64,7 @@ All data models inherit from `Lutaml::Model::Serializable` and use declarative a
   - **Per-record resilience (mirrors the ISO flavor).** A single bad record must never abort the crawl. `#write_file` skips (never `id.content`-crashes on) a record whose MODS entry yields **no docidentifier** (e.g. a DOI stored only in `<identifier type="doi">`, which `ModsParser#parse_doi` doesn't read), and records ids that `#pubid` can't parse (new series like `NIST RB`/`NIST CHIPS`) as "not indexed" — the YAML is still written, only the index-v2 entry is skipped. Both classes accumulate in `#failures`; the `#report_errors` override emits each at `:error` (via `#log_error` → `Util.error`), which the shared `Core::DataFetcher` gh_issue machinery turns into one "Error fetching documents" GitHub issue in CI (`GITHUB_REPOSITORY` + `GITHUB_TOKEN`). `#pubid` no longer warns per-failure; reporting is centralized in `#report_errors`.
   - **Draft-stage rows come from CSRC (relaton#313).** The MODS collection has no draft records, so `#fetch_csrc` adds the CSRC pubs-export feed (`PubsExport.instance.data`) after the MODS pass. It builds each row **exactly as the runtime builds a CSRC hit** — `HitCollection#hit_data` (which calls `#pubs_export_id`) → `Scraper.parse_page` — so the crawled id and the live id are one string, with the stage, revision and update from the row's uri and DOI (`NIST FIPS 197/Upd1 ipd`, `NIST SP 800-108r1 ipd`, `NIST SP 1800-12 2pd`). Only a row whose id carries a **stage** is written; finals come from MODS, so a CSRC row never shadows a base document. On the 914-row fixture feed this writes 302 staged records. Don't go back to a hand-built `"NIST #{series} #{docnumber} #{stage}"` id: it dropped the numbered drafts (`2pd`…), the `/UpdN`, and spelled revisions `Rev. 1`. A stage that CSRC does not list (e.g. `NIST FIPS 197 fpd`) still answers through the consumer's #303 family fallback. Three more rules: `#csrc_bib` clears the `fetched` date the runtime stamps, or every draft changes each day and `Core#force_write` rewrites ~300 files per incremental crawl; CSRC lists a republished draft twice (`…/ipd` and `…/ipd-(1)`, one DOI, so one id), so each id is written once and the plain uri wins; and only the build step is rescued (to a `#failures` entry) — a write error stops the crawl, as in the MODS pass. `#pubs_export_id` must not mutate its row (`sub`, not `sub!`): the crawler and the runtime share the `PubsExport` data.
   - **A CSRC relation names the related document (relaton#314).** `Scraper#related_id` finds the related feed row by its uri and derives its id with `HitCollection.pubs_export_id` (the class method the instance method delegates to). The entry's own `docidentifier` drops the update and the part, and the hit's code is the record's own id, so neither is the related id. `HitCollection#search_filter` ends in an `ensure` that returns `arr`: every branch must set `arr`, because an earlier `return` is overridden there.
+  - **CSWP comes from CSRC, not the NIST Library (relaton-nist#104).** The NIST Library announced CSWP data, but its MODS release (`allrecords-MODS.xml`, and the MARC `allrecords.xml`) holds **no** CSWP record — 0 on the Sept2026 release. It publishes CSWPs only as RIS/BibTeX files under `bib/` on its `nist-pages` branch. So the index has only the CSRC draft CSWP rows (`#fetch_csrc`), and a final CSWP answers from CSRC at runtime: `HitCollection#search` asks the CSRC feed (`from_json`) and asks the index (`from_ga`) only when CSRC gives no hit. That is the chosen rule — CSRC wins, no merge — because CSRC lists every numbered CSWP with more metadata (affiliations, status, supersedes, comment period). The known gap: the NIST Library also has about 35 **legacy dated ids** (`NIST CSWP 01142020`, the same paper as `CSWP 9`, from before NIST renumbered the series) and translations (`NIST.CSWP.6.spa`). CSRC lists neither, so they answer not-found; a fix needs a mapping built from those RIS files. `hit_collection_spec.rb` pins the CSRC answer.
   - **`#write_file` serializes before it indexes**, so a record that fails to serialize leaves no index row naming a file that was never written. That is how a `Bib::Date` in `Scraper.fetch_dates` hid: `Nist::ItemData` takes `Nist::Date`, and YAML serialization rejects the parent class (the same strictness as the docidentifier note below).
 - `ModsParser` — maps MODS XML (via `loc_mods`) to `ItemData`
 - `Scraper` — fetches items from GitHub YAML or CSRC JSON
@@ -118,6 +119,11 @@ NIST-specific gotchas (why this isn't a straight copy of IEC/CCSDS):
   `@raw_content` before wrapping `@pubid` in pubid's `AllParts` — matching
   IEC/CCSDS. `#pubid` itself, read directly, now answers `all_parts? ==
   true` and renders WITH pubid's generic "(all parts)" marker.
+- **The internal form drops `NIST `/`NBS `** (relaton#49), so
+  `docid.to_s(form: :internal)` is `SP 800-113` for `NIST SP 800-113`. The rule
+  is the `"NIST"` row of `Bib::Docidentifier::FORM_PUBID`: `@pubid`
+  renders with `to_s(with_publisher: false)`. A DOI docid has type `DOI` and is
+  unchanged. `to_s` itself is inherited from `Bib::Docidentifier`.
 
 ### Serialization Round-Trip Pattern
 

data/lib/relaton/nist/docidentifier.rb

@@ -74,10 +74,6 @@ module Relaton
         original_content
       end
 
-      def to_s
-        content.to_s
-      end
-
       # `#exclude` (no args) clones the whole graph (incl. a supplement's
       # base), since `walk_chain` mutates it in place. `@raw_content` freezes
       # the stripped rendering — `content` is live from `@pubid`, and the

data/lib/relaton/nist/pubs_export.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 require "singleton"
 
 module Relaton
@@ -61,7 +62,7 @@ module Relaton
       #
       def unzip
         Zip::File.open(DATAFILE) do |zf|
-          JSON.parse zf.first.get_input_stream.read
+          Relaton::Json.parse_external zf.first.get_input_stream.read
         end
       end
     end

data/lib/relaton/ogc/data_fetcher.rb

@@ -1,8 +1,8 @@
 # frozen_string_literal: true
 
+
 require "fileutils"
 require "faraday"
-require "json"
 require_relative "../ogc"
 require_relative "scraper"
 
@@ -126,7 +126,7 @@ module Relaton
         resp = Faraday.new(ENDPOINT, headers: h).get
         case resp.status
         when 200
-          json = JSON.parse(resp.body)
+          json = Relaton::Json.parse_external(resp.body)
           block_given? ? yield(resp[:etag], json) : json
         when 304 then []
         else raise Relaton::RequestError, "Could not access #{ENDPOINT}"

data/lib/relaton/omg/scraper.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-require "json"
+
 require "mechanize"
 require "pubid"
 
@@ -118,10 +118,10 @@ module Relaton
         script = @doc.at_xpath('//script[@type="application/ld+json"]')
         return unless script
 
-        node = JSON.parse(script.text).find { |e| e.is_a?(Hash) && e[LD_DATE] }
+        node = Relaton::Json.parse_external(script.text).find { |e| e.is_a?(Hash) && e[LD_DATE] }
         value = node && node[LD_DATE].first["@value"]
         value && ::Date.parse(value)
-      rescue JSON::ParserError, ::Date::Error
+      rescue ::JSON::ParserError, ::Date::Error
         nil
       end
 

data/lib/relaton/plateau/data_fetcher.rb

@@ -1,3 +1,4 @@
+require "yeptris"
 require "json"
 require "relaton/core"
 require_relative "../plateau"
@@ -78,7 +79,7 @@ module Relaton
         body = hadle_response(response)
 
         # Parse the JSON response
-        JSON.parse(body)
+        Relaton::Json.parse_external(body)
       rescue StandardError => e
         # Handle any errors during the fetching process
         Util.error "Error fetching JSON data from #{url}: #{e.message}"

data/lib/relaton/un/hit_collection.rb

@@ -1,7 +1,8 @@
 # frozen_string_literal: true
 
+require "yeptris"
+
 require "faraday"
-require "json"
 require "securerandom"
 
 module Relaton
@@ -28,11 +29,11 @@ module Relaton
       def search_api(text) # rubocop:disable Metrics/MethodLength
         resp = connection.post("search?l=en&rid=#{SecureRandom.uuid}") do |req|
           req.headers["Content-Type"] = "application/json"
-          req.body = JSON.generate(search_body(text))
+          req.body = Yeptris::JSON.dump(search_body(text))
         end
         return unless resp.success?
 
-        body = JSON.parse(resp.body)
+        body = Relaton::Json.parse_external(resp.body)
         return unless body["status"] == 1
 
         body.dig("body", "data")

data/lib/relaton/version.rb

@@ -1,3 +1,3 @@
 module Relaton
-  VERSION = "3.0.1".freeze
+  VERSION = "3.0.2".freeze
 end

data/lib/relaton.rb

@@ -22,6 +22,7 @@ module Relaton
   autoload :Logger, "relaton/logger"
   autoload :Equivalence, "relaton/equivalence"
   autoload :Core, "relaton/core"
+  autoload :Json, "relaton/json"
   autoload :Index, "relaton/index"
   autoload :Bib, "relaton/bib"
   autoload :ThreeGpp, "relaton/3gpp"

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: relaton
 version: !ruby/object:Gem::Version
-  version: 3.0.1
+  version: 3.0.2
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-10-29 00:00:00.000000000 Z
+date: 2026-10-30 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: addressable
@@ -128,14 +128,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 2.7.0
+        version: '2.7'
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 2.7.0
+        version: '2.7'
 - !ruby/object:Gem::Dependency
   name: faraday-net_http_persistent
   requirement: !ruby/object:Gem::Requirement
@@ -248,20 +248,6 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: 0.1.6
-- !ruby/object:Gem::Dependency
-  name: json
-  requirement: !ruby/object:Gem::Requirement
-    requirements:
-    - - "<"
-      - !ruby/object:Gem::Version
-        version: '3'
-  type: :runtime
-  prerelease: false
-  version_requirements: !ruby/object:Gem::Requirement
-    requirements:
-    - - "<"
-      - !ruby/object:Gem::Version
-        version: '3'
 - !ruby/object:Gem::Dependency
   name: loc_mods
   requirement: !ruby/object:Gem::Requirement
@@ -428,7 +414,7 @@ dependencies:
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: 1.3.112
+        version: 1.3.114
     - - "<"
       - !ruby/object:Gem::Version
         version: '1.4'
@@ -438,7 +424,7 @@ dependencies:
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: 1.3.112
+        version: 1.3.114
     - - "<"
       - !ruby/object:Gem::Version
         version: '1.4'
@@ -460,30 +446,36 @@ dependencies:
   name: pubid
   requirement: !ruby/object:Gem::Requirement
     requirements:
-    - - "~>"
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: 2.0.1
+    - - "<"
       - !ruby/object:Gem::Version
-        version: 2.0.0.pre.alpha.62
+        version: '2.1'
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
-    - - "~>"
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: 2.0.1
+    - - "<"
       - !ruby/object:Gem::Version
-        version: 2.0.0.pre.alpha.62
+        version: '2.1'
 - !ruby/object:Gem::Dependency
   name: yeptris
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: '0.4'
+        version: 0.6.51
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - ">="
       - !ruby/object:Gem::Version
-        version: '0.4'
+        version: 0.6.51
 - !ruby/object:Gem::Dependency
   name: relaton-render
   requirement: !ruby/object:Gem::Requirement
@@ -1149,6 +1141,7 @@ files:
 - lib/relaton/jis/scraper.rb
 - lib/relaton/jis/structured_identifier.rb
 - lib/relaton/jis/util.rb
+- lib/relaton/json.rb
 - lib/relaton/logger.rb
 - lib/relaton/logger/CLAUDE.md
 - lib/relaton/logger/README.adoc