The "IBM 32" characters

IBM Japanese double-byte EBCDIC includes several kanji which do not exist in, or do not round-trip from, JIS X 0208. These were included as gaiji in extensions to Shift JIS and EUC-JP from IBM (e.g. code page 942), NEC, the Open Software Foundation, and Microsoft (e.g. Windows code page 932). However, they were not used as a source for the original Unified Repertoire and Ordering (URO). Instead, 32 of the IBM extension kanji, those which had not been included in the URO from other sources, were included in the CJK Compatibility Ideographs block in the range U+FA0E–U+FA2D.

Of these 32 characters:

• 19 are unifiable with characters in the URO, and are therefore compatibility ideographs in the strict sense.
• 12 are kokuji characters which are actually unified ideographs (with the Unified_Ideograph property, and which do not change upon normalisation). In spite of their inclusion in the CJK Compatibility Ideographs block and their algorithmically generated character names beginning with "CJK COMPATIBILITY IDEOGRAPH", they are not duplicates of characters in the original CJK Unified Ideographs block in any respect;^[6][7] 11 of these 12 are completely non-duplicate, while U+FA23 﨣 CJK COMPATIBILITY IDEOGRAPH-FA23 was later unintentionally duplicated in CJK Unified Ideographs Extension B as U+27EAF 𧺯 CJK UNIFIED IDEOGRAPH-27EAF. They are placed there because they do not have a URO encoding, yet IBM 32 is one of the encodings where duplicate encodings are of concern. All of them are rarely used or are variants of common kanji. They are as follows:

• Uniquely, (U+FA20 蘒 CJK COMPATIBILITY IDEOGRAPH-FA20) is intended to be encoded as the kyūjitai form of a kokuji which received a separate encoding for a variant that is straightforwardly the (extended) shinjitai form U+8612 蘒 CJK UNIFIED IDEOGRAPH-8612. The URO only encoded the shinjitai form, and uses its stroke count to place it in this position. It is furthermore one variant of the many variants of the jinmeiyō kanji U+8429 萩 CJK UNIFIED IDEOGRAPH-8429 (i.e. Kummerowia). U+FA20 was assigned a normalisation to U+8612, even though the 龜 and 亀 components, while both forms of radical 213, are not usually considered unifiable.^[8]