prize18-9  Exercise 2: Reserved Names
=====================================

Predictions: the naive check (exact match) catches only "admin". The key (NFKC, casefold, drop category Cf) should also catch "ADMIN", the fullwidth spelling,
the trailing zero-width space, the soft hyphen in the middle and the zero-width space inside "root" (soft hyphen and zero-width space are both category Cf).
It will not catch "administrator" (a different word, correctly allowed) or the Cyrillic-o "root": that is a look-alike letter, not an equivalent one.

ex2.py:

  import unicodedata as ud

  def key(name):
      n = ud.normalize("NFKC", name).casefold()
      return "".join(c for c in n if ud.category(c) != "Cf")

  reserved = {"admin", "root"}
  tests = ["admin", "ADMIN", "\uff41\uff44\uff4d\uff49\uff4e", "admin\u200b", "ad\u00admin", "ro\u200bot", "administrator", "r\u043eot"]
  for t in tests:
      print(f"{ascii(t):36} naive={t in reserved!s:5} checked={key(t) in reserved}")

Real output (Python 3.14):

  'admin'                              naive=True  checked=True
  'ADMIN'                              naive=False checked=True
  '\uff41\uff44\uff4d\uff49\uff4e'     naive=False checked=True
  'admin\u200b'                        naive=False checked=True
  'ad\xadmin'                          naive=False checked=True
  'ro\u200bot'                         naive=False checked=True
  'administrator'                      naive=False checked=False
  'r\u043eot'                          naive=False checked=False

All as predicted. The last row shows what normalization cannot do: the Cyrillic o is a different letter. Catching it needs a script check (Exercise 1) or a
confusable skeleton, or restricting the field to ASCII. Real systems often combine all three.