RE: Mississippi Hißes
Shawn.Steele at microsoft.com
Mon Dec 14 07:12:19 CET 2009
IDNA2003 disallows fussy.everytype.com and fußy.evertype.com. If you ignore the "is that good or bad" issue, that's "merely" a technical detail of the standard. Similarly you could register the punycode for fußy.evertype.com, and ignore the IDNA2003 mappings and still exchange it with your friends that agreed to the non-standard (per 2003) behavior. So at that point it's not about "being able to control all of the zones", it depends on whether or not the clients do IDNA 2003 mapping.
In short, this is only a technical problem or concern because we choose it to be so, because we choose not to build lookup bundling of these code points (whether PVALID or not) into the standard. If we did, then individual adopters could choose to ignore it as they willed, like ignoring it in IDNA2003 or allowing lookup of other punycode strings that we don't enable.
I don't mind if the working group decides that it doesn't want to build some bundling form into the standard, but that should be presented as "because we choose not to", not "because we can't".
I conceed that I didn't think of Michael's private use of a subdomain. I meant that legally (real laws, not computer standards), it was earlier stated that, in Germany, trademark law would disallow a company from differentiating from another company's trademark only on the form of the ss/Eszett. If our social system disallows such assigmnents, I postulated that it is illogical for us to enable a technical scenario that the target society wouldn't allow us to use. (Except for the my-server-in-my-domain case, which I hadn't allowed.)
Anyway, I have yet to read Vint's doc (tomorrow I will) about transition, so I'd like to respond to that instead, as that's likely a more productive discussion than me randomizing the group :)
From: idna-update-bounces at alvestrand.no [idna-update-bounces at alvestrand.no] on behalf of John C Klensin [klensin at jck.com]
Sent: Friday, December 11, 2009 9:47 AM
To: Michael Everson; IDNA update work
Subject: Re: Mississippi Hißes
--On Friday, December 11, 2009 10:20 +0000 Michael Everson
<everson at evertype.com> wrote:
> On 11 Dec 2009, at 09:32, Shawn Steele wrote:
>> I'm not a lawyer, but I suspect you can't even register
>> fussy.live.com and fußy.live.com individually.
> I can't register fussy.evertype.com and fußy.evertype.com?
> Says who?
Because I fear that this is still not clear, Michael is making
the point that several of the rest of us have been trying to
make in other ways. The DNS is a distributed administrative
hierarchy. In the general case, no one has any control over
subdomains (whether delegated or not, see Andrew's recent
comment about zone cuts) of evertype.com other than Michael; no
one has any control over subdomains of jck.com other than me; no
one outside Microsoft has any control over subdomains of
live.com; and so on for probably hundreds of millions of other
Each of us sets our own policies. Not only does nothing prevent
Michael from putting both "fussy" and the ACE/Punycode-converted
version of "fußy" into the evertype.com zone, nothing prevents
him from also installing "fußy" as a UTF-8 (or even UTF-16 or
UTF-32) string into that zone, nor even from installing it as a
ISO 8859-1 string. Barring his going to considerable effort to
carefully set up aliases (again, see Andrew's recent note and
understand that there are even more possibilities depending on
what he actually wants to do with the names), each one of those
is --both conceptually and operationally-- labels a completely
separate DNS node with whatever record types he wants to attach
at that node. If you can't register both fussy.live.com and
fußy.live.com individually, that is an issue between you and
Microsoft -- not only do the rest of us not care, but someone,
somewhere, will probably conclude that Microsoft's imposing that
restriction is a market opportunity and offer both of those
labels in some other second-level domain.
While, in theory, the TLD registry -- .com in this case-- could
impose contractual requirements on Michael (or me) about what we
could or could not insert in our zone files, no major TLD
registry has tried to do that in years and years, much less
tried to enforce such rules. Even then, the requirements would
be contractual, not anything imposed by the DNS protocols or
other technical requirements.
>From the standpoint of anything this WG --or the IETF in
general-- can do, the decisions are, like it or not, strictly
There are two actual sources of restrictions. I will leave it
to you to estimate how important they are in the general case
(remember, hundreds of millions of zones):
(1) I manage my own zones. I assume from his comment that
Michael manages, or has very close control over, his. Someone
who has some other entity managing a zone for them, or who
manages it via a web interface to a server they don't control,
may be fairly restricted about what they can put into a zone.
If those interfaces were making tests on their input, I assume
they might block UTF-16 or UTF-32 and maybe even UTF-8 and some
other forms (that blocking might occur either as a design
decision or via careless coding). But those would be
restrictions imposed by the chosen operator, not the DNS. And
registrants for domains can change their operators at any time.
(2) With apologies to Mark's wife and her colleagues, trademark
lawyers do get involved in these situations on a case by case
basis, sometimes deep in the tree. It is plausible to assume,
based on several prior episodes, that, while Michael could
install mickey.evertype.com, minnie.evertype.com,
donald.evertype.com, and goofy.evertype.com, he might hear from
someone demanding that he remove them if that set of domains
started to be widely publicized. But that is a different
problem, it wouldn't prevent him from inserting those names into
the zone, and the odds of its occurring if he installed
fussy.evertype.com and fußy.evertype.com as separate nodes is,
Idna-update mailing list
Idna-update at alvestrand.no
More information about the Idna-update