Re: [idn] UTF-8 / RACE

["Adam M. Costello" <amc@cs.berkeley.edu>]

This message contains two replies, one to Eric Hall and one to Sherin
Alsoudani.

"Eric A. Hall" <ehall@ehsco.com> wrote:

> Some of us are arguing the cost of UTF8-with-ACE is only slightly
> greater than the cost of ACE-only.

I don't claim otherwise.

> Some of us are also arguing that the cost of ACE-now-UTF8-later is 2x
> the cost of UTF8-with-ACE.

I don't claim otherwise.  I made no argument about ACE-only versus
ACE-with-UTF8.  I merely argued that ACE is good and should exist.

> Think about the end-user networks out there with admins who are
> struggling with ASCII-based DHCP, lpd, and just about everything else
> that ACE doesn't help with.

ACE helps with all of them by allowing administrators to put IDNs in the
config files now, before the software has been upgraded, if they have an
urgent need.  There is nothing stopping any software from adding support
for UTF-8 (or any other encoding) in their config files, and I doubt
anyone here would want to discourage that.

As for how those protocols represent domain names on the wire, that's
another question, I don't know, but I think that has less of an impact
on the admins.

> It is highly unlikely that those services will ever be fixed until a
> UTF8 form is developed by us...

As soon as we have fully specified the ACE form, the UTF-8 form is
implicitly defined.

Sherin Alsoudani <sherinalsoudani@hotmail.com> wrote:

> To say that "nothing will actually break" is wrong - this is only at
> the protocol etc. level.  However, at the human level, if the IDN is
> an ACE, and my application does not support the proper display of that
> IDN, then something most definitely WILL break: the human user of that
> application will simply not type in that IDN.

Some (maybe most) people will not type in the ACE, and that can be
considered broken.  But someone who really needs to access that domain
name before their application has been upgraded has the option of typing
in the ACE.  And besides, typing domain names is not the only way to use
them.  Replying to mail will work, clicking on hyperlinks will work,
checking SSL certificates will work, etc.  Without ACE a lot of those
operations would break for a lot of users, but with ACE they completely
work for all users.

> People will have to upgrade their applications.

Yes, with or without ACE, people will have to upgrade their
applications.  But with ACE, upgrading their own applications is
sufficient.  Without ACE, they need to not only upgrade their
applications, but also wait for other people to upgrade the
infrastructure, which is out of their control.

> > The provider might be very slow to upgrade, because 99% of their
> > customers might be English speakers, and the other 1% are just
> > screwed.
>
> This is not true...  The response has been very strong from what I
> know: providers ARE switching to UTF-8...

I didn't claim that all providers would be slow to upgrade, I'm saying
that some providers might be slow to upgrade, and without ACE, users
will not be in control, they'll need to wait for the infrastructure to
catch up with them.

> Another thing: assuming 99% English and 1% non-English in the Middle
> East is completely wrong.

I said nothing about the Middle-East.  What about Arabic speakers in the
rest of the world, where they are a small minority?  I bet they'd like
to access the Arabic domain names when they appear.  What about people
who don't even speak Arabic but like to listen to Arabic music?  The
internet is supposed to be for everyone; it's supposed to bring people
together, not put up walls between them.  Without ACE, a domain that
uses an IDN would basically be saying "foreigners keep out".

> I do not understand your comment of "billions of people" who do
> not know "those characters", where those characters are the native
> language of the people.  This is so patently wrong to me that I
> believe I must have misunderstood you.

I believe you must have.  There are billions of people in the world who
do not know Arabic characters, and whose computers do not support Arabic
characters, but sometimes those people will want to use an Arabic domain
name.  Maybe to hear some music, maybe to send a reply to all recipients
of an email message, one of whom is bilingual and has an Arabic address.

> I personally am not interested in seeing English propagated as the
> default language, as a half solution for multilingal domains.  It is
> impossible to ask us to learn the Roman language simply to satisfy the
> protocol writers.

I am not asking anyone to learn any language.  I am suggesting that
occasionally an Arabic speaker will want to access a Chinese domain,
and before they get around to learning several thousand Han characters,
and before they get around to upgrading their computer to support those
characters, they might find it more convenient to learn 52 Roman letters
(38 if you consider upper/lowercase pairs that look the same) and use
the ACE.

> If we have to wait a little or a lot longer for that to occur, that is
> worth it compared to a short-term solution that requires me to send my
> mother to English spelling school.

I'm not asking your mother to use Roman letters to access Arabic
domains.  You should upgrade her applications so she doesn't see the ACE
when accessing Arabic domains.

AMC