Re: New Version Notification for draft-obispo-epp-idn-00.txt

On 22/12/11 18:48, Klaus Malorny wrote:

> I agree to you, but the language tag(s) could be helpful in the coice of
> possible IDN variants. Take the letter "ö". In German, it is legal to write
> "oe" instead of "ö", so "götter" and "goetter" are considered equal and thus a
> registry could reserve the other variant for the registrant of the first. For
> a different language (maybe Swedish?), the "ö" may exist as well, but there is
> no such rule, the two words are considered different and there is no reason to
> block the other. I am in no way a language expert, but I guess there is a
> similar game with accents and esp. with ideographs in other languages.

That's actually a good point, but how exactly they're treated very much
depends on registry policy. For instance, does the registry implicitly
register ASCII variants (as the .cat registry does) or treat them as totally
separate entities (which is EURid's policy), and if you follow a .cat-like
policy, and say (queue silly example) 'foo', 'föö' and 'fóó' are considered
considered distinct words in the language but the convention is to simply
omit the accents, should registering one of them block the others?

That's a pretty complex question.

> Regarding multiple tags: I personally think it should be legal to mix
> languages and scripts under certain (narrow) conditions.

> Why should a domain
> like "café-mосква.ru" not be possible?
        ^^   ^^^  ^
Because it's open to homograph attacks, as demonstrated above. To avoid that,
you'd have to specify some way of deliminating words within the label (as you
did with the hyphen), but that complicates things an awful lot for very little
gain, and you'd have people complaining that they can't register
"cafémосква.ru", which in their eyes might be equally legitimate.

[Aside: To save myself, I'm going to use the abbreviation 'CGL' to refer to
Latin, Cyrillic, and Greek together.]

Now, script mixing isn't necessarily bad. For instance Latin and Hangeul
codepoints in an IDN is safe because there's no chance of homographs. Mixing
CGL, OTOH, is dangerous because all three contain an awful lot of identical
characters. The problem with the CGL scripts is that they are, in essence,
variations on the very same script. Majuscule Latin characters are
practically the same as the Euboean variant of the ancient Greek alphabet.
Cyrillic started out as a variant on the Greek alphabet that borrowed a
number of extra symbols from its predecessor, Glagolitic, to represent sounds
that didn't exist in Greek but did exist in Old Church Slavonic (Old
Bulgarian). This left us with a situation where three now-distinct scripts
ended up with a lot of identical or at least very similar glyphs.

> I don't know whether such a domain name
> make sense regarding usability, but from a registry viewpoint it shouldn't
> matter.

A registry without any form of codepoint whitelisting wouldn't care, but if
we want to avoid abuse of homographs, the choice is either to require the
CGL scripts not be mixed in the one IDN or have rules for how a label should
be partitioned up. Not mixing these three scripts together is, without a
shadow of a doubt, the least troublesome option.

> At the end the basic question is what needs the extension shall cover. As far
> as I can see, the existing and future gTLDs seem to be mostly bound to the
> visions of ICANN (and this is the "one language" approach), whereas existing
> ccTLDs seem to be rather free in their IDN implementations. So the latter
> could have the need of a different approach, but are they interested in such
> an extension at all?

In my experience so far, ccTLDs in countries that use one or more of the CGL
scripts chose the ghostbusters option: don't cross the streams. The Serbian
registry would be an example of this as they have two national scripts. They
allow IDNs in one or the other, but mixing is not allowed. They do not, to my
knowledge, use any extensions for determining which script is used. I'm open
to correction on this though.

K.

-- 
Keith Gaughan, Senior Developer
PGP/GPG key ID: 3E896381
Blacknight Internet Solutions Ltd. <http://blacknight.com/>
12A Barrowside Business Park, Carlow, Ireland
Registered in Ireland, Company No.: 370845
_______________________________________________
provreg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/provreg