Katakana Middle Dot again (Was: tables-06b.txt: A.5, A.6, A.9)

[Andrew Sullivan]

On Sat, Jul 25, 2009 at 07:11:31AM -0700, Michel SUIGNARD wrote:
> I thing that either the contextual rule should be extremely simple 

It strikes me on practical grounds that "extremely simple" is
preferable if only because the alternative is something that will be
possibly difficult to implement.

For the same reasons that we can't just think about TLDs or near-TLDs
when thinking about registries, we had better remember that every
teeny zone all over the DNS is stuck being able to implement these
rules.  Even if we think the right thing to do is, "Don't add it
unless you know what you're doing," we know full well that it's not
going to work like that, and there are going to be zones where
well-meaning but clueless admins admit things that they shouldn't.

If we make very complicated rules (i.e. the sort of thing that can
actually only be reliably tested by a machine), then problems are more
likely, I think.

Given that one of our design goals is to make something permissive
enough to be useful in most contexts, I can see no way to make the
rules just right to allow everything in just the appropriate contexts.
An AI that can pass the Turing test in every human language is
probably not what we want to be designing.

A


-- 
Andrew Sullivan
ajs at shinkuro.com
Shinkuro, Inc.