At 15:13 +0100 3/4/09, Stephane Bortzmeyer wrote:
But this criteria is easier written than applied. For instance, should
the ABNF allow fully-numeric top-level domain names? There is no
*technical* reason to ban them.
There is one.
I am not the best person to describe the entire problem but I'll give it a try.
When it come to IDNs there is a case of "BiDi" (bidirectional) characters.
Some characters have encoded in them the
placement of the next character, to the right or
left. These characters are strong. Other
characters do not have this information, they are
weak.
Numbers and punctuation are weak. Other symbols
(it's not accurate to call them letters) are
strong (not exclusively). Because you might mix
numbers and digits in a domain name (www.163.com)
the weak characters will inherit the directional
information of the other characters.
For the raw data I learned this from, look at these links:
http://stupid.domain.name/node/681
http://stupid.domain.name/node/682
http://stupid.domain.name/node/683
And the only confirmation I have that I'm "on to
something" is this in private correspondence with
the Patrik. I include his comments because I am
still not competent on this subject.
At 1:55 +0100 12/2/08, Patrik Fältström wrote:
#On 1 dec 2008, at 22.06, Edward Lewis wrote:
#
#> So, when pressed to explain this to someone, is this more or less accurate?:
#>
#> With IDNs and specifically bidirectional text (referring to left to right
#> reading languages like Arabic), the direction in which the characters are
#> to be rendered ("is this one to the right or the left of the previous") is
#> sometimes stored in the data of the character itself.
#
# I would say "...is dependent on the directionality properties of the
# character".
#
#> Characters with this data are "strong" and those without are "weak."
#
#"..with explicit directionality data are..."
#
#> Because of the special role of punctuation and
of digits, they are both weak.
#
# "...both weak which implies the order depends among other things on the
# overall directionality of the text."
#
#> Because a dot and a digit are weak, there's no way to guarantee that
#> ".1" will never be seen as "1." in a domain name. There is a way to
#> guarantee it if you include directional aids, but this is apparently not
#> done where domain names appear.
#
#Correct.
[In case you wonder, yes, I'm for allowing fully-numeric TLD, _in that
I-D_, section 3 still being applicable.]
Curious - still?
--
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Edward Lewis
NeuStar You can leave a voice message at +1-571-434-5468
Never confuse activity with progress. Activity pays more.
_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop
