Re: [idn] Reordering evaluation result by existing Japanese JPdomain names

[MIYAYAMA Takayuki <miyayama@nic.ad.jp>]

Dear All,

# This is on behalf of Mr. Yoneya as he is not available now.

On Tue, 13 Nov 2001 14:59:53 -0800
Paul Hoffman / IMC <phoffman@imc.org> wrote:

> >>  Your table are great, and they show that the reordering for Japanese
> >>  names will shorten typical Japanese ACE names by only a tiny fraction.
> >
> >20% is a tiny fraction?  (For all-kanji labels, except the short ones,
> >where compression doesn't matter anyway.)
> 
> Yep. If you look at Yoneya-san's chart, you will see that the large 
> majority of the names are 8 characters or less. In fact, since you 
> want to narrow it to all-kanji names, the large majority are 4 
> characters or less. So, a 20% length reduction means ACE strings that 
> are about 2 characters shorter for typical all-Kanji Japanese names.
> 
> I truly don't believe that we will see end users typing in ACE names, 
> but even if you do, is it really easier to type in h4xe90wsie3b than 
> h4xe90wsie3b5a? This is worth the hair and incompleteness of the 
> reordering draft?

The following are the opinions of those in research of reordering for
Japanese JP domain names.

- It may have some benefit (max: 25% compression) for all-kanji labels
  (39% of the total of JP domain names in Japanese). However, the
  outcome for kanji-kana and kanji-ascii labels which is the rest 61%
  of the total, would amount to max: 16% compression only.

- Within the range of 4 to 5 characters, which is the majority of JP
  domain names in Japanese, even done in all-kanji labels, the benifit
  would remain to be only 2-4 octets.

Considering the overhead of reordering, we don't think it is the
advantage for Japanese Domain Names.

---
MIYAYAMA Takayuki (miyayama@nic.ad.jp)