BIDI rules

[Erik van der Poel]

On Fri, Sep 5, 2008 at 12:56 AM, Stephane Bortzmeyer <bortzmeyer at nic.fr> wrote:
> On Thu, Sep 04, 2008 at 08:41:55AM -0700,
>  Erik van der Poel <erikv at google.com> wrote
>  a message of 30 lines which said:
>
>> Bidi and spoofing are both *display* issues.
>
> Yes. And, therefore, both should be regarded as off-topic for the
> IETF.

No, I think it is reasonable for the IETF to consider bidi display
issues in domain names and to try to come up with rules that can cover
a number of RTL scripts. It seems like we could eventually settle on a
set of rules that would avoid bidi confusion, and that it is
reasonable for the IETF to try to avoid such confusion.

My concern is that the complicated areas like bidi and ZWJ/ZWNJ are
holding up the rest of IDNA200X and Unicode 5.1 in domain names. Many
of the characters introduced between Unicode 3.2 and 5.1 have no
bidi/ZW issues, and many of the IDNA changes from 2003 to 200X are
uncontroversial.

Are we going to keep discussing bidi and ZW until we have perfect
specs for them, or can we publish the uncontroversial parts of
Protocol and Table as Proposed Standards, with informative references
to Bidi and ZW, which would be published as Experimental RFCs?

If we are willing to live with relatively relaxed rules for Latin,
Cyrillic, Greek and Han in IDNA200X, and to leave the visual spoofing
issues in those scripts to zone admins and apps, why can't we live
with relaxed rules for bidi in IDNA200X, leaving the experimentation
to zone admins and apps?

I'd also prefer not to let Rationale hold up Protocol/Table. It seems
like we have consensus on most parts of Protocol/Table, so it doesn't
matter that different individuals would give different rationales, as
long as they agree on the end result.

However, John seems to want to cater to people that would read
Rationale first, therefore requiring normative definitions in that
document. If John is the only WG member that feels this way, perhaps
the WG consensus could be to move the normative parts of Rationale to
Protocol and Table, and publish Rationale some time after Protocol and
Table. Is it time for the Chair to gauge the level of consensus in
this area?

Erik

> On the wire, BIDI labels have zero ambiguity.
>
>> If we can leave spoofing out of the protocol, why can't we leave
>> bidi out of it?
>
> Do you mean "We should regard BIDI display as off-topic and therefore
> stop the work on draft-ietf-idnabis-bidi" (more or less my opinion) or
> "We should forbid RTL characters from IDNAbis", as some people have
> understood your message?
>
>