Minutes from the Tuesday, March 23rd, 2009 IDNAbis WG meeting

[Eric Brunner-Williams]

Hi Mark,

I had to keep myself from running to the mic about this point in time. 
Accessibility (in W3C parlance) to me is remembering to select drupal 
themes that aren't SEO optimized, but content first, sidebars later (if 
at all), and all images used for navigation with ALT tags -- the target 
reader is mechanical translation of text to speech. I've two handicapped 
children, and while their handicaps are not visual, it keeps me thinking 
about what Accessibility really is ... and we used to have a very 
competent person helping with the RFC editing process for whom dial-up 
voice was TTY ... so ...

I don't assume that visual similarity of characters (homoglyphs) is a 
complete statement of the problem, see "MD - suspicious is homoglyphs, 
so "looking" is the test", from my minutes. Not only is "visual" beyond 
the reach of some individuals, and possibly unnecessary for 
text-to-voice implementations, but even "visual" is beyond the reach of 
all but a very small set of people, when the strings are sufficiently 
long. I can't tell the difference between pi to 61 digits and pi to 62 
digits, and as a kid, I actually watched my parents to crash dump 
analysis with character-width rules to find corrupted memory (mid-1960s, 
IBM and CDC mainframes). No body does that kind of work anymore, and 
damn few people can find the "S" at offset 30+/- in either of those two 
representations of pi.

I don't think Vint made the best counter-argument to your position, 
which I'm glad you restated, for a search engine which creates 
identifiers intended only to be processed by algorithms, not eyeballs, 
but as Chair that's not something he should do anyway.

I don't think that the only type of "suspicion" that is of concern for 
us is visual confusability, though I'm not trying to insert "sounds 
alike" into the "looks alike" mess, I just think the exclusive focus on 
glyphs isn't certain to deal with all character problems.

If credit cards and instant resolution weren't presumed truths, our 
notion of what constitutes "suspicious" would likely be very different, 
as well as its apparent importance.

So, now I've done my little lazy evaluation run to the mic.

Eric

Mark Davis wrote:
> Thanks to Eric as well; taking notes in that kind of meeting is very 
> hard. For the one case where what I said was "missed by scribe", my 
> response was something like:
>
> There are different kinds of "suspicious" domain names; for example, a 
> company like Google keeps list of known spoofing sites. However, 
> /that/ kind of suspicious domain name has nothing to do with the 
> character content of the URL, nor with the choice of characters to 
> allow in domain names. The only type of "suspicion" that is of concern 
> for us is visually confusability.
>
> Mark