Refutation of listed arguments for 3-character minimal length of IDN gTLDs

[Werner Staub <werner@xxxxxxxx>]

The Explanatory Memorandum "Discussions about the 3-char String
Requirement" of contains a series of mistaken arguments in support
of requiring 3 "distinct characters" for each gTLD string. (Page 5 on
http://icann.org/en/topics/new-gtlds/three-character-30may09-en.pdf)

I submit a refutation of each of these. For easer reference, I copy each
mistaken argument as quoted text followed by the refutation.

As all the arguments brought forward to date are false, the proper way
forward is to adapt the Draft Applicant Guidebook. One solution is to
make the minimum length dependent on script. ICANN should also allow the
respective language communities to request a minimum string length based
on considerations specific to the underlying languages or script.


REFUTATION OF ARGUMENT 1: "Fairness of treatment"

The following statement is false:

 "In addition to comments received from the CJK community, ICANN
  received arguments from the European region that certain single or
  two character combinations in European languages represent a word
  or a meaning and in some cases also geographic identifiers. These
  arguments were made to counter arguments for allowing less than 3
  character strings per the proposal above. If less than 3 character
  strings are allowed for CJK based languages then they should also
  be allowed for other languages, for fairness of treatment."

Proposing "equal" treatment in terms of uncomparable measurements grossly
unfair. By analogy, imagine a food rationing system where a set amount of
food were alloted for to each building, irrespective of how many people
lived there.

To make it even clearer, imagine for a moment that the DNS had always
been in Chinese ideographs, and that we Westerners were asking for ASCII.
How would we feel if we were told that the domains should be at least 3
separate words? Or if we were told that all characters used should be
composed of a least 5 separate brush strokes? Or that a TLD should
require at least 6 keystrokes? (Most Chinese and Japanese two-letter
words require over 7 keystrokes to type.)

The fact that a small number of word exists in European languages that
are composed of just characters (or one) does no change this. In any
non-ideographic language, there are very few words composed of just 2
characters. By contrast, most words in CJK that expressing a generally
understood concept are composed of just two ideographs, and at least 1000
often used Chinese and Japanese words are just one character. As for
Korean, words _appear_ to a Westerner as one or two characters, but in
reality these are syllable blocks composed of multiple Jamo characters.



REFUTATION OF ARGUMENT 2: "Statements concerning the Chinese words and
number of characters"

The following argument is false:

 "Other arguments state that few Chinese characters are words, most
  Chinese words are two or more characters.  If one separates out
  the specific phonetic implications (one character equals one
  phoneme), characterizing Chinese characters as syllables would be
  much more accurate. In addition, some opinions are that most
  Chinese words consist of more than one character."

In Chinese almost all characters correspond to one syllable, but this
does not change the fact that each ideograph has a meaning as a
word. Most ideographs can be used alone and in combination. It is only
logical that there are more word combinations than single words: in this
respect all languages share the same feature. In Japanese, each ideograph
has at least 2 different readings and generally at least one of them is
pronounced as two or more syllables. In Korean, ideographs are rarely
used today as most writing is in Hangul, a phonetic script. The latter
indeed appears as blocks of syllables. However, each block that appears
as a Hangul "character" to a Westerner is actually a block of two or more
individual phonetic characters.

The overwhelming majority of words one can find in a Chinese, Japanese or
Korean dictionary are composed of what appears as one or two "distinct
characters" to a Westerner. These are actually compound words, i.e.
combinations of full words. It is impossible to see how one could justify
the imposition of a Western-style "abbreviation" by requiring more
characters than the full word!



REFUTATION OF ARGUMENT 3: "Trial of a few gTLDs with less than 3
characters"

The discussion document reports the following suggestion:

 "Some suggestions have been made that ICANN perform a trial
  implementation of a certain small number of gTLDs that have less
  than 3 characters. This would then be used to inform the
  development of the process for allocating such strings more
  widely."

There are no stability issues to address with such a trial. Requiring
such a trial would be a deliberate and malevolent act of discrimination.



REFUTATION OF ARGUMENT 4: "Translations of TLDs"

The following argument is fundamentally misguided:

 "Comments have been received that in relation to translation of
  existing TLDs, there has never been a model for "translating"
  TLDs. Therefore, the two-letter ISO codes or other TLDs that are
  “abbreviations” cannot be translated into IDN strings of less
  than 3 characters as a meaningful representation of that
  definition. They are not standardized abbreviations and
  abbreviations are not a standard concept across languages and
  cultures. The ccTLDs in particular are a standardized coding
  system, chosen as codes for a number of reasons including
  recognizability and distinctiveness of undecorated Latin
  character."

There is no reason why IDN TLDs should be "translations" of Western (or
any other) expressions, and there is no need for them to be
"abbreviations" in the Western sense.

The proposed Chinese TLD .公司 (gongsi) means "business" or
"corporation". It is present in most registered company names in China.
It is neither a translation of .com, nor a translation of .biz. And it
should not be. It is a full word, more than an abbreviation. If two
characters give us a complete word (remember, they require 6 keystrokes
to type), why would anyone pretend that it should be lengthened to 3
characters?

About language-to-language translations of TLDs (or other names), I might
add that they tend to be either hilarious or confusing. They could hardly
be of serious use. A generalized translation-based approach to TLDs is
utterly impossible.



REFUTATION OF ARGUMENT 5: "ICANN’s ccTLD delegation function"

The following concern does not exist:

 "Currently the IANA delegation function relies on the scarce
  availability of 2-character ASCII combinations and all of these
  (when entered in the ISO3166-1 list) are treated as ccTLDs.
  Discussions in the ICANN community as well as at ISO MA meetings
  in the past few years have focused on the feasibility of expanding
  the ISO3166-1 list to contain 2-character combinations of other
  scripts, representing country and territory names. This would
  require a multi-year ISO led development (which might occur after
  the ccNSO IDN PDP). The outcome of the Fast Track Process will
  inform the ongoing discussion about whether or how to expand the
  ISO3166-1 list and associated ICANN ccTLD delegation function, as
  well as the long-term ccNSO PDP for IDN ccTLDs. Delegation of
  single and two-character labels now, might jeopardize the future
  shape of the ccTLD delegation mechanism."

It is natural to extend the ISO-3166-1 list with corresponding country
*names* and *well-known abbreviations* in all languages. But names and
abbreviations are not *codes*. It would be a very bad idea to create more
than list of standard codes for the same things.

The ISO3166-1 list is of codes is based on the basic 25-letter Latin
alphabet. This is the only character set present in all the computer
systems and all education systems around the world. This means that
ISO3166-1, as a list of codes, is addressable in all systems around the
world. There is no need for a translation of a code into news codes.
There is a need for translation from code to *names* in specific
languages. The ISO-3166-1 list is being updated in this respect and will
eventually contain country names and abbreviation (not codes) in all
major languages and in the respective scripts.

It must also be stated clearly that it is not possible to "translate" a
code from one script to another. Translation is possible from language to
language and from code to language. For instance, it is possible to
translate the name "USA" from English to "EEUU" in Spanish or to 美国
(měiguó = "beautiful country") in Chinese or "米国" (beikoku - "rice
country") in Japanese. The code "US" can thus also be translated into
each of these language-specific names. But it there is not way to create,
for instance, a "code" that translates "US" into ideographic script: any
such attempt would lead to confusion.

There can thus only be mapping of ISO-3166-1 from code to all names in
all major languages (as opposed to a mapping from script to script).
Names cannot be restricted to 2 "letters". It has already been shown that
IDN ccTLDs cannot be limited to 2 letters. For instance, the ideographic
name of Singapore, 新加坡 (Xīnjiāpō) cannot be shortened to 2 characters.

The mapping codes to names and well-known abbreviations in major
languages is a compilation effort. It is not standardization, as the
country names exist independently of the standards or documentation
authority. This compilation effort is already protected in the gTLD
process as country names may only be used with the approval of the
relevant government.

As a result, there is no imaginable standardization or compilation
effort whatsoever that could possibly be "jeopardized" by allowing
2-character or 1-character ideographic gTLDs.

Werner Staub