[Comments-korean-lgr-25jan18] Open Letter to the Proposal for a Korean Script Root Zone LGR
Shinjo Park
peremen at gmail.com
Tue Feb 13 16:14:31 UTC 2018
Authors: Jaemin Chung, Seonghoon Kang, Shinjo Park
Dear whom it may concern,
This is an open letter to the Proposal for a Korean Script Root Zone
LGR, published as
https://www.icann.org/en/system/files/files/proposal-korean-lgr-25jan18-en.pdf,
made in public on 2018-01-25. We also attached a PDF copy of this
letter in this message.
We are deeply concerned about the recent proposal for allowing both
Hangul and Hanja in the Korean script (as opposed to Hangul only) in
internationalized domain names (henceforth the Proposal). We believe
the support will give negligible benefits to the Korean community
while increasing the potential to confuse. It will also make
similar-looking names open for exploit, and introduce new
accessibility issues. We believe that the Proposal should be
withdrawn.
# Exaggerated Hanja Usage in Korea
The Proposal claims that Hanja usage is significant in Korea to
warrant its support in IDNs. This is greatly misleading and we
consider the evidence listed in the appendices of the Proposal to be
intentionally chosen to exaggerate its usage. In this section, we will
review them and show why they are not representative of what IDNs will
encode.
## Coexistence, Not Substitute
> Today, Korean is written in Hangul. _Hanja are sometimes
provided in parentheses next to Korean words,_ but only when the word
in Hangul alone may be misunderstood due to its multiple meanings or
when further clarification of a specific meaning is necessary. _In
rare cases, words are written in Hanja and the particles and suffixes
associated with the words are left in Hangul._
(Emphases ours)
It is true that Hanja still enjoys the continuous usage as a
supplementary script in the modern Korean language. But its usage has
been greatly reduced to clarify homophones in Hangul and independent
usages are much rarer, even mentioned in the Proposal itself. Unless
IDNs allow annotation strings inside a label, the mere coexistence as
a supplementary script is not a good fit for IDNs.
## Registered Trademarks That Do Not Warrant Current Use
The Proposal shows multiple mixed-script trademarks seemingly
registered in recent decades. This is misleading, as their origins
could be traced back to several decades ago when the proprietor of
trademarks actively used Hanja-based ones. While they are no longer
actively used, the trademark renewal can be explained as a defensive
measure against expiration or misuse for older trademarks that are
irrelevant to current usage.
For example,
* Samsung Moolsan (more commonly known as Samsung C&T Corporation)
This particular trademark traces back to at least 1975
(registration number 4000484540000). Of course, Samsung C&T no longer
uses the mixed-script trademark today; it instead uses a trademark
common to the Samsung group and a logotype for “삼성물산”.
* Samsung Jeonja (more commonly known as Samsung Electronics Co., Ltd.)
This particular trademark traces back to at least 1985
(registration number 4001346350000). Again, this is not the current
trademark used by Samsung Electronics.
* Hyundai Motors
This particular trademark traces back to at least 1969
(registration number 4000191560000).
* Kumho
This particular trademark traces back to at least 1986
(registration number 4100077620000).
* Taehwa Shopping
This particular trademark traces back to at least 1983
(registration number 4100043980000). The company itself went bankrupt
in 2001 and renamed to Judies Taehwa in 2003.
In most recognizable cases like Samsung and Hyundai, it is not hard to
see that these companies no longer use Hanja trademarks in Korean
publicly. One can argue that Hanja trademarks are used in Chinese and
Japanese markets; we will later iterate that supporting Korean Hanja
only is not sufficient for Chinese and Japanese speakers.
## Decorative Nature of Hanja
It is not as prominent in signboards (due to the selective nature of
examples), but many mixed-script trademarks are common in that the
only portion written in Hanja is 주식회사 (株式會社, lit. _corporation_). This
is not a coincidence. Being a script taught in school but rarely
written, Hanja has acquired a supplementary role to decorate some
words. We have chosen the word “decoration” for this purpose because
its usage is limited to common words, like 남 (男, lit. _male_), 대 (大,
lit. _big_; often short for _university_ [대학교]), 가 (家, lit. _home_).
If they were meant to be emphases, any word of importance would be
written in Hanja.
We believe that such decorations are not to be written but to be only
read, and an IDN label is not a place to reproduce decorations.
## Other Bad Examples
Remaining examples comprise corporate registers and law books, both
highly atypical as examples of a mixed-script IDN. The authors of the
Proposal seem to have failed to find more convincing evidence of
Hanja. We sympathize with the authors as it will be quite difficult.
In reality, Hanja usage is so low that neither Android nor iOS
provides Hanja input methods in Korean by default. Please note that
this is not just an oversight; smartphone penetration in Korea has
reached 90% of the total population and remains extremely high across
all ages, as observed in Gallup Korea’s 2017 survey [1] for example.
If Hanja cannot be readily typed what’s the point in allowing Hanja
IDN labels?
## Hanja Will Be Used in Parallel with Hangul in Primary School Textbooks?
Appendix H.6.2 cites an article from The Hankyoreh [2] mentioning that
Hanja will be included in primary school textbooks, starting from
2019. However, later article from the same newspaper [3] dated 9th
January 2018 states that the inclusion of Hanja in primary school
textbooks will not be implemented. Given that the Proposal was
published on 25th January 2018, two weeks after the announcement, the
validity of the entire section of the appendix needs to be questioned.
# Potential Ambiguities
One may still recognize even the slightest use of Hanja and argue for
its support in IDNs. We believe that, however, in addition to low
usage, mixed-script IDNs will be actively harmful to the overall
Korean community, due to confusion not only within Hanja, but also
between Hangul and Hanja.
## Hangul-Hanja Confusion
The Proposal lists five variant groups of similar-looking Hangul and
Hanja, only two of them in the repertoire. This is a massive
underestimation and we believe that there can be tens or even hundreds
of them; it is so well known that it got a name Yaminjeongeum (야민정음)
on Korean Internet, a pun using the name of the supposed origin and
the original name of Hangul, Hunminjeongeum. Namuwiki, a
community-driven encyclopedic wiki, has a great list [4] of them.
Below is the non-exhaustive list of Hanja characters included in MSR-2
which have the similar visual appearance with Hangul characters
(either jamo or syllable), which are not included in the Proposal’s
section 6.2:
* 勹 (U+52F9)↔ㄱ (U+1100)
* 廿 (U+5EFF)↔ㅂ (U+1107)
* 刁 (U+5201)↔ㅋ (U+110F)
* 大 (U+5927)↔ㅊ (U+110E)
* 刀 (U+5200)↔ㄲ (U+1101)
* 丗 (U+4E17)↔ㅃ (U+1108)
* 从 (U+4ECE)↔ㅆ (U+110A)
* 金 (U+91D1)↔숲 (U+C232)
* 長 (U+9577)↔튽 (U+D2BD)
* 辛 (U+8F9B)↔푸 (U+D478)
* 卒 (U+5352)↔쭈 (U+CB48)
* 卫 (U+536B)↔고 (U+ACE0)
* 告 (U+544A)↔솜 (U+C19C)
* 丕 (U+4E15)↔조 (U+C870)
* 奀 (U+5940)↔좆 (U+C886)
* 笑 (U+7B11)↔쑻 (U+C47B)
* 豆 (U+8C46)↔묘 (U+BB18)
* 号 (U+53F7)↔묵 (U+BB35)
* 吴 (U+5434)↔뭊 (U+BB4A)
* 彐 (U+5F50)↔크 (U+D06C)
Even with a simple search on the Internet, we can easily find
look-alike characters which are not listed in the Proposal. Given the
fact that this list is non-exhaustive containing less than 1% of
characters included in MSR-2, we are afraid that there are much more
pairs unknown to us.
## Same-Script Confusion
Chinese characters are generally noted for their visual complexity,
which suggests higher potential of same-script confusion.
The complexity of Hanja can also contribute to phishing attacks. Let’s
take an example of 검찰청 (檢察廳, Prosecutors’ Office). The Hangul spelling
검찰청 is used daily, while some news headlines use the initial
character’s Hanja equivalent 檢 as a symbol. However, all three
characters in its Hanja spelling have quite large numbers of strokes,
making them hard to distinguish especially in low-resolution and/or
small display. If Prosecutors’ Office wants to register its Hanja-only
IDN, it would be 檢察廳.한국. Assume that an attacker manages to register
檢祭廳.한국, which shares the first and third characters while only the
second one is different (察 (U+5BDF) vs. 祭 (U+796D)). Unlike Chinese
and Japanese speakers who use Hanzi or Kanji every day, a
substantially high number of Korean speakers will believe 檢祭廳.한국 is
also the correct IDN since the full Hanja spelling is not used on a
daily basis.
Another example is 청구서 (請求書, invoice). The Hanja spelling 請求書 is not
used daily, and unlike the example of 검찰청, none of the Hanja
characters are used as a symbol. Suppose that legitimate companies
could use their name with 請求書 suffixed as a Hanja IDN for their
electronic invoice system. If an attacker wants to phish as an
electronic invoice system, they can buy the Hanja IDN 請求晝. This IDN
shares the first and second characters with the legitimate one, while
only the third one is different (書 (U+66F8) vs. 晝 (U+665D)). Unlike
the Prosecutors’ Office example, even a partial Hanja spelling is not
used daily, therefore making users more vulnerable to same-script
confusion.
In fact, Korea already paid a lot of social resources to counter
domain-name-based phishing even without IDNs. Prosecutors’ Office [5]
and invoices [6], listed as examples here, can easily become real
victims of domain-name-based phishing. An extra countermeasure will
have to be added to defend against Hanja-based phishing.
Although chances of same-script confusion also exist in Chinese and
Japanese IDNs, due to the low daily usage of Hanja in Korean, it will
create much more confusion when Hanja is introduced into Korean IDNs.
## Multiple Representations of Mixed-Script IDNs
When a name consists of multiple Korean words, there can be multiple
combinations of Hangul and Hanja for that name. Let’s take an example
of 대전도시공사 (Daejeon City Corporation, doing city infrastructure
management like German Stadtwerk plus real estate development). This
name is comprised of three individual words: 대전 (大田, city of Daejeon),
도시 (都市, city), and 공사 (公社, state-owned enterprise). In Hangul-Hanja
mixed-script IDNs, these three words can be written in either Hangul
or Hanja. All of these domain names are possible: 大田都市公社.한국,
大田도시공사.한국, 대전都市公社.한국, and so on. Unless all possible domain names are
taken by the legitimate owner, someone else can register any possible
domain names that are not taken and pretend that they are the real
corporation. To avoid such a problem, owners of multi-word domain
names need to register all possible combinations of Hangul-Hanja
mixed-script IDNs, or only register the one that is likely to be most
commonly used and hope for the best. The difficulty of choosing one
from all possible combinations can be a detriment to safety and
discourage the usage of Korean IDNs altogether.
# Other Problems
We have found other problems in the Proposal, which further weaken the
validity of the entire proposal.
## Incompleteness of Repertoire
The K portion of IICore comprises 4744 (not 4743) characters, which
consists of KS X 1001 (without 14 characters) and additional 138
characters. According to Dr. Ken Lunde [7], an expert on CJKV
information processing, the details of how these 138 particular
characters were selected are unknown. He also says that there is no
document or report explaining how the K portion of IICore was
prepared, and the people who compiled the K portion of IICore either
passed away or are no longer participating in the Korean National
Body. It is questionable why the Proposal should include IICore
characters.
## Accessibility Issues
We believe that IDNs in general need to be platform-neutral. To this
end, Korean IDNs are not meant to be used only on Microsoft Windows
computers with the KS X 5003 Korean keyboard. Appendix H.5 in the
Proposal shows how to enter Hanja with the Korean IME on Microsoft
Windows. According to StatCounter [8], there are about 15% of users
who are not using Microsoft Windows as their PC operating system. Even
though widely used input method applications in macOS and Linux
support Hanja input, the Proposal fails to show PC operating systems
other than Microsoft Windows. We believe that the Proposal should have
included Hanja input methods for all these three major PC operating
systems for the sake of completeness.
Although the Korean Standard KS X 5003 includes dedicated keys for
Hanja conversion and Hangul/Alphabet toggle, not all keyboards sold in
Korea are equipped with those keys, especially for laptops where they
are shared with the right Ctrl and Alt keys respectively. Similarly,
Apple keyboards sold in Korea do not have them at all [9].
Moreover, web browsers are not only installed on PCs today. But the
Proposal also fails to address Hanja input on non-PC environments.
There are other media devices with web browsers such as smartphones,
set-top boxes, smart TVs and game consoles. They are more
resource-constrained than a PC, and Hanja input requires a big
dictionary like Chinese and Japanese input methods. Designers usually
leave only Hangul input functionality in Korean locale to avoid the
overhead introduced by Hanja data. While Hanja input requires
candidate selection – which could be time-consuming without efficient
candidate navigation interface – Hangul input methods can directly
translate QWERTY or ten-key keyboard keystrokes to final precomposed
characters, making Hangul input more efficient than Hanja input in an
environment without the traditional PC keyboard. For those reasons, it
is impractical to implement Hanja input in addition to Hangul input on
every such device. Given increased penetration of Internet-enabled
devices without Hanja input, chances of utilizing Hanja IDNs on such
devices will be very low. Presenting Hanja input only on Microsoft
Windows PC is not a proof that Hanja input is universally available
for all Korean speakers.
Domain names are also used in spoken language. Alphabet-only and
Hangul-only domain names have no or minor difficulties in spoken
language. In contrast to both types of domains, dictating Hanja in a
spoken conversation is hard. People may use different approaches to
describe Hanja, such as
* meaning (훈) + reading (음),
* word-based (e.g., the character A in the word AB), and
* shape-based (e.g., component C on the left and component D on the
right; or component E which looks like something).
But all of them are inefficient and inconvenient in a spoken
conversation. Besides, even if we ignore the fact that these are
inefficient and inconvenient, they cannot always work (note that the
examples below are all from KS X 1001).
* A single combination of meaning and reading can correspond to more
than one Hanja. For example, “클 석” can mean either 奭 (U+596D) or 碩
(U+78A9); “옥돌 민” can mean either 玟 (U+739F) or 珉 (U+73C9).
* There may not be well-known words. Sometimes there might be no
word at all. For example, 贇 (U+8D07) is not used in any Korean word
(only appears in personal names).
* There are characters whose shapes cannot easily be described, such
as 寡 (U+5BE1), 秉 (U+79C9), and 肅 (U+8085).
Also, when a listener has no background knowledge on Hanja, or the
domain name consists of an uncommon or ambiguous Hanja character, it
will be much more difficult to reconstruct a Hanja IDN from a spoken
conversation.
To make things even worse, Hanja IDNs will greatly inconvenience
visually impaired users, as they need to be read aloud by a screen
reader or presented in braille. Korean screen readers can read aloud
Hanja inside plain text or in a Hanja candidate window of an input
method application by only its reading or its meaning + reading,
depending on the user preference [10]. If a Hanja IDN is read aloud
just by its reading by a screen reader, it is not possible to
distinguish whether the given IDN is using Hangul or Hanja without
changing user preference. Even when both meaning and reading are read
aloud by a screen reader, the ambiguity issue mentioned in the
previous section still applies.
While there are two unofficial incompatible Hanja orthographies for
Korean braille [11] (there is no “official” supervision of denoting
Korean Hanja in braille according to the 2017 Korean braille
orthography [12]), the report states that the demand for Hanja braille
among visually impaired Koreans is low. As a reflection of this low
demand, a Korean braille terminal named 한소네 (Hansone) developed by
Selvas Healthcare only supports Hangul and Latin alphabet input
without Hanja input capability [13]. Unlike Chinese and Japanese,
Hangul-only IDNs require no processing of Hanja – which is inevitable
in mixed-script or Hanja-only IDNs – on screen readers and braille
displays. This Hanja processing problem can be avoided by not
introducing Hanja into IDNs in the first place.
# Information Exchange with Chinese and Japanese Speakers?
One possible argument for allowing Hanja in Korean IDNs is information
exchange with Chinese and Japanese speakers. In fact, Hangul-Hanja
mixed-script IDNs and IDNs with Korean Hanja will likely cause
confusion to Chinese and Japanese speakers rather than helping them
due to lexical and regional differences.
Let’s take the city of Seoul as an example. Owing to its etymology,
Seoul cannot be written in Korean Hanja. Due to this fact, Seoul has
exonyms 汉城 (Simplified Chinese Hanzi) and ソウル (Japanese Katakana).
Given the Chinese exonym caused confusion within Korean speakers, the
city proposed 首尔 as a transliteration of the Korean pronunciation into
Simplified Chinese in 2005. When encoding a domain name of Seoul City
Hall, one hypothetical mixed-script IDN could be 서울市廳.한국. This
actually causes confusion to Chinese and Japanese speakers: 1) they
can neither understand nor enter Hangul ‘서울’, ‘한국’; 2) Chinese and
Japanese use different words for “city hall” – 市政府 and 市役所
respectively.
If you still don’t grasp lexical differences, let’s take a look at
this Chinese word: 足球. If 足球 or its Korean reading (족구) is presented
to Korean speakers, they would have no idea that this means “soccer,”
as 족구 (足球) in Korean is a domestic sport distinct from soccer
(“soccer” in Korean is 축구 (蹴球)). You cannot blindly assume that
anything written in Chinese characters will be understood with the
same meaning regardless of languages. It is not always true.
If Korean speakers want to target Chinese and Japanese speakers, they
need to register IDNs in Chinese (首尔市政府 and 足球) and Japanese (ソウル市役所
and サッカー), not the ones in Korean written in (Hangul and) Hanja (서울市廳
and 蹴球).
Chinese and Japanese speakers input non-Korean variants of Chinese
characters. Even though Chinese characters are called CJK “Unified”
Ideographs in Unicode and ISO/IEC 10646, there are lots of characters
separately encoded due to the Source Separation Rule. The same
abstract shape is sometimes unified and sometimes separately encoded.
Let’s take a look at these examples: 友情.한국 and 靑年.한국 (these are
commonly used words). Chinese and Japanese speakers will have no
problems with inputting 友情, but will face a problem when inputting 靑年.
Why? This is because ⿰忄靑 and ⿰忄青 are unified (情 U+60C5) but 靑 and 青
are separately encoded (靑 U+9751 vs. 青 U+9752) due to the Source
Separation Rule (靑 is used in Korean, and 青 is used in Chinese and
Japanese; this also applies to 情). There are lots of examples like
these (e.g., 僧 unified, 增/増 separated; 愉 unified, 兪/俞 separated; 慨
unified, 槪/概 separated; 朗 unified, 郞/郎 separated; etc.). Ordinary
users are not aware of the Source Separation Rule and don’t know which
ones are unified and which ones are separated.
We believe that Korean Hanja domain alone is not helping information
exchange with Chinese and Japanese speakers.
# Conclusion
We believe that the Proposal has to be withdrawn, given the technical
and practical flaws. Although Hanja is used as a supplementary script
in the Korean language, its usage does not simply fit into what IDNs
need to encode. The Proposal misrepresents the usage of Hanja and why
it should be coded into Korean IDNs. It also underestimates possible
confusion between Hangul and Hanja and within Hanja, especially for
Korean speakers. Mere addition of Hanja in Korean IDNs will introduce
new accessibility issues, especially for visually impaired users. We
believe that the benefit of using Hanja in Korean IDNs is much lower
than the harm caused by it.
# References
[1] http://www.gallup.co.kr/gallupdb/reportContent.asp?seqNo=813.
2017-02-15, Retrieved 2018-02-12.
[2] http://www.hani.co.kr/arti/society/schooling/776746.html.
2016-12-30, Retrieved 2018-02-12.
[3] http://www.hani.co.kr/arti/society/schooling/827055.html.
2018-01-10, Retrieved 2018-02-12.
[4] https://namu.wiki/w/%EC%95%BC%EB%AF%BC%EC%A0%95%EC%9D%8C?rev=1559#s-2.7
야민정음 (r1559 판) – 나무위키 (Yaminjeongeum (revision 1559) – Namuwiki).
2018-02-12, Retrieved 2018-02-12.
[5] http://www.ilyoseoul.co.kr/news/articleView.html?idxno=211304 –
The article mentions that the phisher gave a fake website of
Prosecutors’ Office. 2017-11-14, Retrieved 2018-02-07.
[6] http://blog.alyac.co.kr/1353 – Figure 2 in the article shows an
example of a phishing email with Apple’s invoice, which redirects to a
fake website containing “Apple” in their domain name. 2017-09-28,
Retrieved 2018-02-07.
[7] https://blogs.adobe.com/CCJKType/2018/02/exploring-iicore-part-1.html.
2018-02-05, Retrieved 2018-02-07.
[8] http://gs.statcounter.com/os-market-share/desktop. 2018-01,
Retrieved 2018-02-07.
[9] https://support.apple.com/en-us/HT201794. 2017-06-06, Retrieved 2018-02-07.
[10] https://www.freelists.org/post/nvda-korean/NVDA-1 NVDA용 한자 데이터
1차로 완성했습니다. (Finished an initial work on Hanja data for NVDA) – the
email mentions that the NVDA data file contains multiple formats of
read-aloud Hanja in Korean text. 2012-09-08, Retrieved 2018-02-07.
[11] http://www.korean.go.kr/common/download.do?file_path=bookData&c_file_name=e5ab7da6-8b09-4f2f-9d6d-595d8e9ea1e6_0.pdf&o_file_name=%ED%8A%B9%EC%88%98%EC%96%B8%EC%96%B4-01-10.pdf&downGubun=bookDataView&book_seq=312
한자점자 규정 제정에 관한 기초연구 (A preliminary research on Korean Hanja braille
orthography). 2012-12, Retrieved 2018-02-07.
[12] http://www.korean.go.kr/common/download.do?file_path=etcData&c_file_name=5b0b6d52-0375-4fc8-98f0-912b26c0ee47_1.pdf&o_file_name=2017%20%ED%95%9C%EA%B5%AD%20%EC%A0%90%EC%9E%90%20%EA%B7%9C%EC%A0%95.pdf
2017 Korean braille orthography. Retrieved 2018-02-07.
[13] http://www.himsintl.co.kr/download/hasone5_manual_170628.docx 한소네
5 사용자 설명서 (Hansone 5 user manual). Retrieved 2018-02-07.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: letter.pdf
Type: application/pdf
Size: 294111 bytes
Desc: not available
URL: <http://mm.icann.org/pipermail/comments-korean-lgr-25jan18/attachments/20180213/bc8b7fd5/letter-0001.pdf>
More information about the Comments-korean-lgr-25jan18
mailing list