Re: [idn] hostname history hell

[John C Klensin <klensin@jck.com>]

James,

Neither Dave nor I used the word "useful".  I carefully avoided
it and he probably did to.

It is a reality of human nature that, for almost any idea, if
you go around and ask enough people "would this be useful" or
"might you find this attractive to have", without any sense of
the tradeoffs or costs, someone will say "yes".  Probably even a
lot of them will say "yes", possibly accompanied by "why not".  

_I_ can think of a number of reasons why it would be really neat
to have arbitrary symbol combinations in domain names. I'm not
very imaginative about these things, and certainly am not
formulating marketing plans about how I can make more money
"selling" more names, which might motivate me to get more
imaginative.  But, for example, I can think of all sorts of nice
uses for a domain named @#$%^.org, URIs notwithstanding.

But...

The primary job of the WG is not to figure out a way to include
anything that someone might think is useful.   The job is to
incorporate international character identifiers into the DNS (I
hope we all understand by now that "names" is as much a misnomer
as "multilingual") while being carefulto not wreck the Internet
or the DNS.  That "while" inherently implies tradeoffs and I
believe that we should confine ourselves to the clearly-needed
(and ideally lowest-risk) cases.  That is a much higher standard
than "someone thinks this would be useful" or "there is market
demand".

Remember, as a corollary to Milo's principle (*), we know how to
make pigs fly.  The fact that we don't do it often is the result
of an engineering tradeoff: it is too expensive and the
mechanism is unacceptably hard on the pig.

     john

* Milo Medin, while at NASA, pointed out that anything could be
made to fly given sufficient thrust.


--On Wednesday, 21 November, 2001 16:53 +0800 "James
Seng/Personal" <jseng@pobox.org.sg> wrote:

> The problem is we cannot determine what is useful and what is
> not.
> 
> For example, SGNIC may decide that it only allows Han
> Ideograph, Tamil scripts and US-ASCII only. JPNIC may decide
> otherwise. Perhaps some registery in future things
> 'symbol-drawing' is useful (e.g. a registry for trademark).
> 
> I suggest we follow the simple-and-dumb rule for IDN.
> 
> idn -> any ucs characters subjected to nameprep prohibition
> (section 5) ihn -> ...registry define themselves...
> 
> application can use nameprep to "check" idn.