Re: [ga] IDN issues (was: On Elections)

[JFC Morfin <jefsey@xxxxxxxxxxxxxxxx>]

At 07:13 06/10/2007, Sotiris Sotiropoulos wrote:
> JFC wrote:
>
> > (6) IDNs, and further on LDNs (local/language Domain Names), raise a
> > large number of issues, in which none of the ICANN numerous
> > committees have discussed or engaged in proper dialog with the
> > concerned experts.
>
> I see no reason why subscribers to the GA cannot voice their opinions
GA Members are not accepted in ICANN IDN committees: these are 
reserved to Constituency members. In addition some internal disputes 
may not help (I keep refering to Danny and I over the NCUC Membership 
of my organization, because it is the only one which is a little 
known on the GA [not because of any feud against Danny :-)], but 
there are many others ones which are ignored).
> openly and candidly about their preference for "multilingualization" as
> opposed to "internationalization" of the IDNspace, you have done an
> admirable job of stating your concerns and preferences.  Thank you for
> sharing them with us...  Personally speaking, while I sympathize with your
> conception of the "multilingualization" of the IDNspace,
This is not really mine.
It is a world consensual requirement.
Now, the necessary standardization paradigm is confirmed.

> I am not sure
> that the IAB and the IETF have managed to work out all the problems of an
> English based ascii "internationalizing" standardization model,
After years of work, one sees that a fully successful 
internationalization is not possible in the TCP/IP model. There are 
many reasons as to why, but the blocking reason is the lack of a 
presentation layer in the "DoD Model". As a pure non-secure e-system 
(English inside), the Internet architecture did not need it - saving 
on its prototyping complexity and cost. Now, this layer is to be 
faked at the user application level, using different hybrid solutions 
such as XML, IDNA, semantic web, etc.
There are three possibilities:
- inserting a presentation layer. This is a huge thing to do, but I 
think of something.
- universally faking one (as we are faking it on a per 
application/SSDO [like W3C] basis)
- building a virtual Internet atop the current one which would include one.

If we do not want to waste time or work for a long time, I think this 
should be optimized with the result that IETF will come up with (or 
not) the addressing and routing. This is a major difference, for 
example, if we have an IPv6 Internet or "IPv8" Internet.
> so to
> introduce untold numbers of new complexities to the mix by trying to phase
> in a multilingual approach, should that even be possible at this point,
> may actually cause more confusion than benefit...
Actually, multilingualization is much simpler to technically consider 
than internationalization (we only know this via the confirmation 
that the failed Debbie's NWIP brings, as to how the world wants it, 
and this is what has an impact. The same, the IETF has not yet 
explored John Klensin's proposition.
This being said, you are correct, IAB/IETF, Open Sources, W3C, etc. 
have only had internationalization in mind for more than a decade, 
and you cannot change that. This is why internationalization must 
continue to develop and stay abreast of multilingualization's 
parallel emergence. There is still a lot of work to be made at the 
internationalization layer, but not as in excluding the 
multilingualisation layer any more.
To understand this, one has to understand internationalization and 
multilingualisation are not at the same layer in language support 
model and we have to correct the current IETF layer violation. 
Therefore, all this should mutually simplify and assist. There are 
five communication layers to support languages:
(1) no worded language.
This is universalization by numbers, signs, signals, music, etc.

(2) lingualization.
You embed a language into your model. This should only be at the 
application layers. This is not the case for the Internet, which only 
has a protocol pile.
(3) globalization (fathered by Mark Davis, current Unicode President).
You increase (a) the language script support (Unicode), (b) the end 
ability to support some other language (localization), (c) use a 
common reference and filtering grid (langtags).
If you do it at the network layer, you "lingually bias" the Internet, 
and if you do it at the application layer, you "lingually bias" the 
application.
(3.1) This is why they accepted IDNA (Internationalized Domain Name 
Application) as there was no architectural problem. The problem is 
that ICANN does not want to manage it as an application layer issue 
(using xn--abcd formated regular DNs and TLDs), but rather as a 
network issue, as if the IDN were actually run at the network level. 
This is complicated because IDNA is Unicode based. And Unicode is 
definitely no part of a stable, crisp, secure network layer system as 
ASCII is. There are solutions to that, but the main problem is to get 
people to accept that there is a problem (look at the IPv6 situation today).
(3.2) The whole BCP47 issue that I had at the IETF is that they 
documented the second approach (applications), attempting to make 
people believe that it was the first one (network) for marketing 
reasons (access to IANA). They did not want to discuss IDN because 
they were an _other_ application and not a protocol.
(4) multilingualization
This is what I described before. Every language can be supported 
equally on the technical side of things (obviously real support will 
depend on the way people use the tools - English, Chinese, and French 
are likely to obtain more supporting content than Inuit). To show 
that there is no contradiction between globalization and 
multilingualization, a good definition of multilingualization is "the 
localization of globalization", which means to globalize every 
language. This seems to be a huge and impossible task to centralizing 
American globalizers. Therefore, they block themselves at their 
globalization/internationalization layer. Debbie brought them the 
correct decentralizing way of an English globalizer, which should 
dramatically help them to organize their transition, but she still 
confused it with a generalised distributed solution.
(5) metalingualization
The reason as to why they think it is complex is because they are 
starting from one language (English, but it could be Chinese, or any 
other single language), so they have no rule/method/know-how/model. 
They would proceed with a second language the same as with the first 
one, in turn creating conflicts. This is what is feared regarding 
what could come from China and Chinese Names.
If you change paradigm (the way everyone considers something) you can 
more easily understand how all this works. But which one? As 
indicated above, the vote shows that ISO prefers to work on its 
existing distributed paradigm rather than on the decentralised one 
Debbie proposed; but there is interest in seeing the evolution that 
Debbie's proposition could bring.
What is the current ISO paradigm? A paradigm is also obtained from an 
example that may serve as a model. ISO 3166 is the paradigm that 
shows we can agree to write (in languages and scripts) about 
languages and scripts independently from languages and scripts. This 
is the expression of a more general axiom: metadata semantics is 
language and script invariant. This ultimately means that semantic 
decoders will allow you to speak in your language and for me to 
receive _your_ semantic in mine (actually better than I do today 
without such an assistant and through the pragmatic screen - I need 
to know all the context). This can be achieved in at least in two ways:
- using polynym (synonyms in every language) solutions such as 
translation memory,
- or conceiving conceptual languages such as IEML. (Canadian Pierre Levy).

John Klensin's major contribution is to have suggested to only 
consider scripts in the IDNA process and not to feel bound by any 
linguistic limitation. This means that scripts would belong to the 
IETF, not languages. Scripts are digitally coded information, 
languages are the codes for semantics.

Practical implementation
Internationalization and IPv6 are much alike: they are adapted to the 
evolution of the Internet from the rough consensus unilateral 
approach of the architecture towards a decentralized influence (this 
is the IETF framework that was recently detailed by Harald Alvestrand 
in RFC 3935). This can work for an international US network (it was 
invented at IBM to keep in close touch with their foreign users). 
Multilingualization is like relativity: atoms did not change when 
Einstein discovered it, but they were considered differently.
Therefore, as it is for IPv6, and the current ROAP problem (the IETF 
meets with routing and addressing), the language problem is the way 
that they consider the existing technology, resources, and users.
Languages are a key issue, but only one of similar key innovative 
issues that lead to better perceive the Internet as distributed. 
Actually, the IETF relation with Internet machines, is the ICANN 
relation with us, the users of these machines.
The same way ICANN wants to impose on us their old perception of 
users as people that can be led through influence, the same way IETF 
is tied to their old perception that the Internet is a decentralized 
place (this is a basic IETF core value) that can be influenced. They 
do not consider distributed spaces where interoperability and 
interintelligibility is the key. Both of them are to get real. The 
internet is not only made of ICANN contracts and old RFCs.

Impact
The point you make is, therefore, rather important. Languages are a 
main part of externets axiologies:
- externets can be called "external network look-alike within the 
network", "colsed gardens", "user groups", etc. they are not very 
well known to Internet architects. John Klensin proposed to use 
classes to support the Multilingual Internet that way (ICANN - ICP.3)
- axiologies are set of parameters and functions that may define a 
virtual environment [ex. a standard] (that for example an ontology 
may decribe [ex. a norm].
Multilingualization de facto brings up this concept (the English 
externet, the French and Greek externets where only one language is 
used). In order to implement externtets, other concepts will emerge , 
such as "maps" (the real-time description of externet topologies). 
This makes us enter into considerations such as NIMROD [a key IETF 
project in terms of understanding the Internet], but on a multilple 
basis (a global Internet wide map for each externet). Now multi-tier 
is legal in the US, externets would offer an excellent cross-ISP 
architecture etc.
However, at the same time, externets means a big loss of legitimacy 
and technical capacity in controlling the "externets of the Internet 
of local nets" (networks of the network of networks). Practically, 
this means that the only network common end to end continuity 
everything should be built upon would be "interoperability". This is 
totally conform with the "protocol on the rope" and "dumb network" 
Internet principles, but not with the way ICANN tries to govern it, 
IETF has permitted black boxes to develop and is late with OPES, and 
ISO/IANA have not cooperated yet.
This means that interoperability can only be obtained today through 
an Open Standard and Open Governance approach. No more IETF 
standards, US dominance, ICANN control, or even ITU - only an 
Interoperability Governance Forum and ISO norms (as explained above: 
a norm describes "normality" [average situation] used as 
interoperable bottom-up standardization [a standard describes the way 
one wants to change normality]).
Can this work?

It has to. We have to remember that we, as persons, can stay alive 
and do something else if we become bored, computers cannot: if they 
are poorly used they just stop delivering and the network dies.
Cheers.
jfc

"First they ignore you, then they laugh at you, then they fight you, 
then you win!" Gandhi.