Use a two-part data format version in IndexedDB metadata.

content/browser/indexed_db/leveldb_coding_scheme.md

 # LevelDB Coding Scheme
 
 LevelDB stores key/value pairs. Keys and values are strings of bytes,
 normally of type `std::string`.
 
 The keys in the backing store are variable-length tuples with
 different types of fields, described here using the notation «a, b, c,
 ...». Each key in the backing store starts with a ternary prefix:
 «database id, object store id, index id». For each, the id `0` is
 always reserved for metadata; other ids may be reserved as well.
@@ skipping 132 matching lines @@
 - - - -
 
 ## Global metadata
 
 The prefix is «0, 0, 0», followed by a metadata type byte:
 
 key                                 | value
 ------------------------------------|------
 «0, 0, 0, 0»                        | backing store schema version (Int) [`SchemaVersionKey`]
 «0, 0, 0, 1»                        | maximum allocated database (Int) [`MaxDatabaseIdKey`]
-«0, 0, 0, 2»                        | SerializedScriptValue version (Int) [`DataVersionKey`]
+«0, 0, 0, 2»                        | data format version (Int) [`DataVersionKey`]
 «0, 0, 0, 3»                        | primary BlobJournal [`BlobJournalKey`]
 «0, 0, 0, 4»                        | live BlobJournal [`LiveBlobJournalKey`]
 «0, 0, 0, 100, database id (VarInt)» | Existence implies the database id is in the free list  [`DatabaseFreeListKey`] - _obsolete_
 «0, 0, 0, 201, origin (StringWithLength), database name (StringWithLength)» | Database id (Int) [`DatabaseNameKey`]
 
 *** aside
 Free lists (#100) are no longer used. The ID space is assumed to be
 sufficient.
 ***
 
+The data format version encodes a `content::IndexedDBDataFormatVersion` object.
+It includes a 32-bit version for the V8 serialization code in its most
+significant bits, and a 32-bit version for the Blink serialization code in its
+least significant 32 bits.
 
 ## Database metadata
 [`DatabaseMetaDataKey`]
 
 The prefix is «database id, 0, 0» followed by a metadata type Byte:
 
 key                    | value
 -----------------------|-------
 «database id, 0, 0, 0» | origin name (String)
 «database id, 0, 0, 1» | database name (String)
@@ skipping 140 matching lines @@
 the same user (index) key in non-unique indexes prior to the inclusion
 of the primary key in the key itself. `0` is always written now
 (which, as a VarInt, takes a single byte)
 
 *** note
 **Compatibility:**
 The sequence number and primary key, or just the primary key may not
 be present. In that case, enumerators that need the primary key must
 access the value.
 ***