relaton 3.0.1 → 3.0.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f81699e1df94f1c570f01c07b4114492b7cc5ae1e45715a4ced3ef51c1ba3fb6
-  data.tar.gz: 1ea38d455d16b512807a505bd558757ea0509a9392cf41c120f3fc10801beb87
+  metadata.gz: a774b326943089257e3aef6e5402fc1e2d53d4ff557e8b19f845a62d64b9a99b
+  data.tar.gz: 665fa378d2380f79a4e2952df1ff3d0f7e1de6d91ead262d75261555a4170e5c
 SHA512:
-  metadata.gz: 4d27294461f66031023053e1ec70a4c26b0c8fa068e6ed9d1224d1ab9c6cb366a2b2d8f85cd50305c4826733ab028817f70b96aebc267346d224b9be1fdf29be
-  data.tar.gz: a3c00ef5a93352f08a492624b531955494c8530a6b0d68f01dc748d58d8599304fa9ac743d157cfd2a6cc37b63664c781da75de5e199de70035ed133e70d0305
+  metadata.gz: bfa2af357a4e288f0ce97fb82ff78ac58b39bc8f8fc8659584ebb29a4568a50175137b8b8cd32b6710aff65b8413cb4ab0d1ea674918187edd9a74592e331a6c
+  data.tar.gz: e2ad6a9822a6ffa5f9a625bb4095260b81b3906fcff8ba0cf4fbdcea0d543f9e2b14133c7dd29dad0f165c26713b16fc0f06ebfc06d134384d4d0f280ee873b1

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/bipm.rb

@@ -1,6 +1,5 @@
 require "zip"
 require "fileutils"
-require "parslet"
 require "relaton/bib"
 require "relaton/index"
 require "relaton/core"

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/processor.rb

@@ -67,7 +67,7 @@ module Relaton
       # @return [Pubid::Identifier, nil]
       def cache_pubid(ref)
         pubid_class&.parse(ref)
-      rescue ArgumentError, ::Pubid::Errors::Error, Parslet::ParseFailed
+      rescue ArgumentError, ::Pubid::Errors::Error
         # relaton-bib carries the log facade; core specs load without it.
         require "relaton/bib"
         Relaton::Bib::Util.info "does not parse as this flavor's pubid; " \

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/registry.rb

@@ -221,7 +221,7 @@ module Relaton
         require "pubid"
         parsed = ::Pubid.parse reference
         parsed if parsed.to_s == reference
-      rescue ArgumentError, ::Pubid::Errors::Error, Parslet::ParseFailed
+      rescue ArgumentError, ::Pubid::Errors::Error
         nil
       end
 

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)) ||
+        false
+    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
@@ -370,7 +389,7 @@ module Relaton
       return unless docid&.content
 
       cache_key docid.content, nil, {}, stdclass
-    rescue ::Pubid::Errors::Error, Parslet::ParseFailed
+    rescue ::Pubid::Errors::Error
       nil
     end
 
@@ -437,7 +456,7 @@ module Relaton
     def copublisher_query(code, opts, stdclass, publisher)
       query = begin
         @registry[stdclass].query_pubid(code, opts)
-      rescue ::Pubid::Errors::Error, Parslet::ParseFailed
+      rescue ::Pubid::Errors::Error
         nil
       end
       return query if query

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
 
       #