🚀 go-pugleaf

In article <2791@titccy.cc.titech.ac.jp>
mohta@necom830.cc.titech.ac.jp (Masataka Ohta) writes:
>In article <1jfgq1INNqmn@flop.ENGR.ORST.EDU>
>	crowl@jade.CS.ORST.EDU (Lawrence Crowl) writes:
>
>>The question I was asking was "can you _identify_ a han/kanji character
>>based on a sequence of radicals"
>
>No, you can't. Radicals are for indexing only. The rest of the character
>has its own complex shape.

If you can use radicals for indexing, then you can use them to identify
characters.  The process of identifying a character need not be able
to, by itself, generate an acceptable image.

>>and "would it be reasonable to encode
>>han/kanji on that basis".
>
>Such encoding is too lengthy.

An encoding every variant of every character ever written is not?  At
214 radicals, we can represent a radical in eight bits, and a character
in 8*(average number of radicals per character).  Your non-unified
approach would require roughly eighteen bits per character.

>>Agreed.  However, there is no natural size for tables.  Table sized of
>>4000 are much cheaper than table sizes of 64000.
>
>If you use radical based encoding, it makes everything complex.

Could you please elaborate?  Your argument leaves me unconvinced.

>Moreover, you will have to have sixteen 4000 entry tables which is as
>large as a single 64000 entry table.

No, I don't have to have sixteen 4000 entry tables.  I only need one.

>>But, can sixteen bits represent _all_ historical Han characters _and_
>>the historical texts of all other languages?  My guess is 16 bits can
>>_if_ Han characters are coded as radicals,
>
>Maybe nor may not be. Many complex Han characters are just unique.

Unique in what sense?  Examples?

>BTW, from the view point of programmers, combining characters are
>just unusable.

I am a programmer.  It is from a programmer's view that I made my
proposal.  What I am proposing is not unusable.  In many ways it is
more usable that straight character coding.

-- 
  Lawrence Crowl		503-737-2554	Computer Science Department
			   crowl@cs.orst.edu	Oregon State University
	      ...!hplabs!hp-pcd!orstcs!crowl	Corvallis, Oregon,  97331-3202