On Fri, 27 Oct 2000 Marco.Cimarosti@... wrote:

> Pierpaolo Bernardi wrote:
> > A secondary benefit of having a canonical decomposition for
> > hanzi is that
> > this decomposition could be used as a base for a radical+strokes based
> > ordering.
>
> This is possible only if the decomposition and the radical-stroke method
> share the same logic, which is probably quite unlikely.
>
> E.g., if you want an ordering based on the traditional Kangxi radicals, but
> the decomposition is based on the writing order of the components, the
> matching between the two systems becomes quite troublesome. The Kangxi
> radical is often the first component, but not always: some radicals are
> normally written after the phonetic, some are written partially before and
> partially (the "sealing stroke") after it.
>
> Moreover, the counting the number of strokes in a component (or in a hanzi)
> may be quite arbitrary, because different fonts may lead to different
> counts.

In addition, one can also use the 214 Kangxi radicals without filing
characters under the same radicals as the AD 1716 _Kangxi Zidian_ (U+5EB7
U+7199 U+5B57 U+5178) dictionary did, which is what some modern
dictionaries have done. In reality, these 214 radicals were used earlier
in the AD 1615 _Zihui_ (U+5B57 U+5F59) and the AD 1671 _Zheng Zitong_
(U+6B63 U+5B57 U+901A), but the _Kangxi Zidian_ was what popularized
them (and the use of the term "zidian" to mean 'dictionary') and
made them a de facto standard, and provides a common ground for
character-using countries from which divergence may occur.

Depending on how close one wants to be faithful to the _Kangxi Zidian_,
one could also use the stroke counts from that dictionary, which matters
to some degree, as it affects the order of the radicals themselves.
However, it presents problems for characters mistakenly left out
at time of compilation or invented afterwards; e.g., a few hundred of
the ~13,000 "traditional" characters in the Big5 character set used in
Taiwan and elsewhere are not in that dictionary.


Thomas Chan
tc31@...