emoji (was Re: I-D Action: draft-klensin-idna-rfc5891bis-00.txt)

Andrew Sullivan ajs at anvilwalrusden.com
Sun Mar 12 23:53:18 CET 2017

On Sun, Mar 12, 2017 at 10:27:32PM +0000, Shawn Steele wrote:
> > The approach that IDNA2008 
> > took was that the DNS did not need to permit any label anyone might 
> > want, but instead needed to permit effective mnemonics that people 
> > could use reliably.
> But that's apparently not how people want to use IDN/DNS.

Unfortunately, the basic problem here is that there are things we
might want that are in fact not technically possible, and the trade
offs are not always self-evident.  Moreover,

> Yup, but those aren't the requirements the users want.

I find it essentially impossible to determine "what the users want",
because I don't know who the class of "users" is here.  Some of the
"users" are basically people who have a commercial interest in finding
new ways to sell registrations of stuff that maybe will appear in the
DNS.  I suspect that, if you tell some "users" (i.e. purchasers) of
such names that the names turn out to be unpredictable in their
behaviour, they might have a different reaction.  The point of at
least the IETF standards is supposed to be interoperability, and the
interop story here is quite bad (whatever UTS#46 has decided to say

> I'm not sure why.  Sure, they aren't great mnemonics, however if you resolve I♥NY.com to an IP address, then, then that server has been identified.

Your argument seems susceptible to the fairly obvious reductio that,
if you manage to arrive at or
2600:3c03::f03c:91ff:fe69:70e2, the server has _also_ been identified.
That an identifier might work in some circumstances hardly shows that
it is a suitable general-purpose identifier for the Internet, which is
what we were talking about.  Moreover,

> ellow sticky note - but then again, I can't tell you how to type a Chinese IDN either.  I might be able to do a barely recognizable attempt at Cyrillic or Greek, but you'd better know what I meant!

these are especially bad analogies _precisely_ because they are
well-developed writing systems where the distinction between
"competent reader and writer of the language" and "incompetent reader
or writer of the language" is reasonably apparent to other competent
language users.  The same is simply not true of emojis, partly because
the context for their use is _exactly_ the sort of informal uses where
ambiguity and imprecision is less important.  Even rendering is
largely undefined.  Indeed, http://emojipedia.org/pile-of-poo/ shows
that at least one vendor doesn't put a happy face on the pile of poo:
is that a different emoji or isn't it?  Those distinctions are what
make emojis perfectly fine for personal communications and really bad
for identifiers.
> Modern use of domain names isn't about finding an IP address.  It's about branding.  That may not be an answer that engineers are comfortable or happy with, but it's how normal people use the Internet.

I think that the "about branding" slogan you claim there is rather
more reductive than even the IP lawyers I keep talking to here at the
ICANN meeting would be comfortable with.  I said to someone only
earlier this evening, we need to develop policies and procedures that
work from multiple perspectives: we cannot build these things with
just one hand.

Best regards,


Andrew Sullivan
ajs at anvilwalrusden.com

More information about the Idna-update mailing list