[UA-discuss] More from Ram Mohan on ICANN's further commitment to Universal Acceptance

[Richard Merdinger]

Thanks for this write-up, John.  I think you stated the complex issues quite succinctly and well.  I agree also that it is not us to UASG to attempt to provide solutions, but that it is important for us to acknowledge the issues and help them gain visibility so the right groups can drive to some viable resolutions.

--Rich

-----Original Message-----
From: UA-discuss <ua-discuss-bounces at icann.org> On Behalf Of John Levine
Sent: December 24, 2018 12:01 PM
To: ua-discuss at icann.org
Subject: Re: [UA-discuss] More from Ram Mohan on ICANN's further commitment to Universal Acceptance

In article <6963BB0C-4172-4EBD-9C38-56DEDFECF45E at lboro.ac.uk> you write:
>-=-=-=-=-=-
>Will ICANNʼs further commitment to UA extend to ICANN having their own set of IDNs and EAI addresses? ...

>I have never understood this reluctance/resistance.

I do.  

Nobody understands what it means for two different domain names to be "the same", or for two web sites to be "the same", or even for the same web site addressed by different URLs to be "the same."  Do all the names redirect to the original URL, or do you have a complete version of the web site at every IDN address?  How do you try to keep bookmarks straight if there are four URLs for every page?  For a site like ICANN's that has multiple language versions of many pages, do you try to make the language in the page match the language of the URL, or do you prefer the Accept: languages from the user's web browser, or what?

It is relatively straightforward to have a Chinese web site at a single Chinese IDN domain, but nobody has a clue how to parallel versions of web sites at different names.  I've done some informal surveys of names that are supposed to be "the same", even in the same language, with names in .cat when they were doing DNAMEs, and names in .ngo and .ong, and I found that hardly anyone even tries to make it work.

For EAI, as my TLD survey showed last month, only about 10% of mail systems for gTLDs can handle EAI mail, with a large fraction of those being hosted at Gmail and Outlook.  From a UA point of view, if you want to communicate with people, it is a good idea to handle other people's EAI addresses, but a poor idea to assign EAI addresses and expect other mail systems to handle them correctly at this point.*

There is a further swamp with addresses used as identifiers, as does approximately every web site in the world that has user accounts.  We have informal conventions for ASCII addresses: upper and lower case are equivalent, and addresses with exotic punctuation don't work very well.  We have nothing like that for EAI, and in view of the vast number of different and incompatible conventions in different languages, it is a very hard problem.  I offred a little informal advice on address assignment in the EAI guide I wrote, and the IETF may do some work in this area next year, but we are a long way from solving or even understanding it.

Finally, I am definiteky NOT suggesting that UASG should try to solve any of these problems. We need to be aware of them and perhaps warn people away from situations that might cause more UA issues.  We barely have the time and expertise to do what we're already trying to do.  Let us avoid mission creep here.

R's,
John

* - This may be different among communities that all speak the same non-Latin language, perhaps in parts of India, but it's true in general.