Comments on IDNA Bidi

JFC Morfin jefsey at
Thu Jan 17 00:50:45 CET 2008

At 19:07 16/01/2008, Michel Suignard wrote:
>Please don't confuse the bidi terminology more than it needs to be.

Dear Michel,
I am afraid my priority is not the bidi terminology, but stable 
domain names to be used by people from every language.

>There is no difference between display and visual order, except that 
>I would agree with Cary that display order is probably a better term 
>and in fact Unicode uses the 'display' term more often than 'visual'.

Well, I suppose I will not make you to change your mind. However, a 
very quick poll around during this evening diner shown me that 
everyone commonly accept there can be a difference between the order 
they see something on their TV screen and the order it was 
progressively displayed.

>However the visual order (or visual representation) term has been 
>used in documents related to bidirectionality for a long time. I 
>came across that term since the late 80s when general purpose 
>Operating Systems I was dealing with started to implement 
>bidirectional handling and it was used frequently by the bidi 
>experts at that time (mostly from IBM) and the logical versus visual 
>order comparison has taken root since then. So in practice 'visual' 
>and 'display' order tends to be used interchangeably.

I understand this, in the legacy printing technology. Also, I 
understand you are considering your own ISO 10646 code.

What I see is that these solutions are unsucessful in order to 
address my multilingual DNS zone Manager problem. I am no expert in 
your approach. I will not point where you exactly meet difficulties. 
I only see that I have a need delayed for eight years and no 
satisfactory solution offered by IDNA. I identify most of the 
listed/discussed problems as related to the choice of Unicode.

So, I consider what we should have considered in 2001 and what we 
agreed upon when we closed the WG-IDNA, i.e. to consider a Plan IDNB 
if  IDNA met deployment problems. There is nothing special about it. 
It is the same as IETF considered IPv5, IPv6 and IPv7 in parallel 
before chosing to focus on IPv6.

There are two solutions:
- IDNA-200X, i.e. to try to review IDNA-2003 and to constrain 
everything to fit together. I think it would require limitations 
Unicode will not be able to accept at some stage. Because it is not 
designed for that, and has many other applications to support. Also, 
this seems extermely complex and not adequate for the most used 
industry today: naming, accross scores of naming systems which are to converge.
- IDNB, i.e. to use another code set, designed for other 
applications, that could support multilingual naming.

Then to test and compare.

>In addition, what is referred in the message below as 'display 
>order' is unclear to me. What is registered is for sure what 
>everybody else would call 'logical order'.

Not necessarily. This is what I quickly refered to in using "keyboard 
order". I agree it was an over simplification.

>  Unicode is not a standard to just manage what you called 'display 
> order'. See page 19 of the Unicode standard 5.0 for further details 
> concerning logical order and the interaction with readable (displayed) text.

For me "readable" is "visual", but let not dispute on qualifications. 
Let focus on what is the real underlaying difference: the dynamic of 
labels. You are familiar with a static "display" approach, 
considering the end result, with different rules and possibilities to 
achieve that result.

- What you consider as stable is the final displayed result being 
always the same.
- What I consider as stable is the process to be unique and always the same.

There may be cases where ISO 10646 can deliver my stability, with the 
help of a version of Unicode, cases it cannot. Not because it would 
be bad, but because it was not designed for that.

>Keyboard order is another beast altogether as it depends on the 
>input method editor you use and varies a lot among implementation of 
>writing systems input.

See above. I used "keyboard order" as a simplification and to 
underline the dynamic aspect. Time of display related order.

>And please stop to try to put a wedge between ISO 10646 and Unicode. 
>ISO 10646 for the longest time has had normative references to the 
>Unicode bidi algorithm and the normalizations forms and both 
>standards are more and more aligned not only in content and 
>terminology (I guess I should know being project editor for 10646).

The difference I make is between a 2003 ISO user voted table and 2006 
version manufacturer published rules. I also make a difference 
between the ISO and the Unicode organisations nature, basis, trust 
and scope. I doubt the UV3S I need will be worked on by the Unicode 
consortium (or by the JTC1/SC2/WG2), but I am sure it will eventually 
become an ISO international standard.

Why? Let consider what is to be changed in the TC145 charter to 
strictly fullfill the IDNB needs: "Standardization in the field of 
graphical symbols as well as of colours and shapes, whenever these 
elements form part of the message that a symbol is intended to 
convey, e.g. a safety sign. Establishing principles for preparation, 
coordination and application of graphical symbols. General 
responsibility for the review and the coordination of those already 
existing, those under study, and those to be established. The 
standardization of new graphical symbols" what suits me as a users.

I suppose, you will probably use the end "when requested by a 
technical committee, or where it does not fall within the activity of 
an existing technical committee. Excluded:  standardization of 
letters, numerals, punctuation marks, mathematical signs and symbols, 
and symbols for quantities and units. However, such may be used as a 
component of a graphical symbol." and object, delaying the solution 
we need. IMHO you would be correct, permitting ISO to consider what 
is the most efficient oganisation.

However, during that time I would still wait. As you may know, I am 
not ISO. I only know that the WSIS has also identified that kind of 
need. From experience, I also know that internal ISO limitations do 
not necessarily impeach my positions to be supported by ISO votes. 
The real point for me, is therefore not to waste time and efforts of 
my team and to understand what the world (DNS, semantic addressing, 
customs, business, etc.) really need before working on an issue (I 
agree with you) which is not a small one.

I am sure that at some stage an NWIP will be proposed. Then it will 
be a matter of votes and additional help we will certainly need.


>For the foreseeable future there is no alternative to the 
>Unicode/10646 tandem, we can improve and there is room for that, but 
>trying to create a new paradigm does not seem a useful exercise to me.

I take you say this as part of Plan IDNA: I do support you in that 
context, so you make the best out of the IDNA proposition.
I am sure you will beave the same about people working on Plan IDNB.

>Sorry for this slightly out of topic set of remarks.

They would be out of topic if IDNA supporters opposed IDNB. This 
would not be IETF and not efficient.

>-----Original Message-----
>From: idna-update-bounces at 
>[mailto:idna-update-bounces at] On Behalf Of JFCM
>Sent: Wednesday, January 16, 2008 7:20 AM
>To: Cary Karp
>Cc: idna-update at
>Subject: Re: Comments on IDNA Bidi
>At 11:40 16/01/2008, Cary Karp wrote:
> >"Display" strikes me as the more objective term (being an attribute of
> >what the machine does), to which "visual" is a cognitive correlate
> >(being in the mind of the beholder).
>I agree. When you enter a name on a keyboard, this is the display
>order. When you read it on an ad, this is the visual order. Unicode
>is a standard to manage the display order (printing). IDNA wants to
>use it in a visual order (the order used to register it). This
>creates several impossibilities to address the DNS, zone
>administrators, security, etc. needs, including because the same
>visual may correspond to several display codes or even sequences and
>therefore orders.
>The only DNS needs are a unique registration order and a unique
>registration correspondance between display and visual. Bidi is a no
>problem once the direction is set (it results from the display order
>qualias [position and left to right or right to left]).
>This leaves us with two plans:
>- plan IDNA, to best use the current Unicode oriented proposition on
>a temporary basis, that countries and semantic addressing can use now.
>- plan IDNB, to devise a universal visual sign secured set
>interoperable with ISO 10646 and a conversion from that UV3S to ASCII
>that can be consistently used in every semiotic applications, and at
>International Network presentation layer (that we also have to devise
>and implement).
>Why to support the idea of a single authoritative root, so the same
>ASCII DN resolves the same IP everywhere, if the same U-label does
>not resolve first to the A-label for ever?
>Idna-update mailing list
>Idna-update at
>Idna-update mailing list
>Idna-update at

More information about the Idna-update mailing list