Re: [idn] Roundup on optional characters

[Keith Moore <moore@cs.utk.edu>]

> At 4:19 PM -0400 8/15/00, Keith Moore wrote:
> >  > - Some people like the idea of optional characters in scripts where
> >>  there are common use for them. Of the people who actually chose
> >>  between (a) [optional], (b) [prohibited], and (c) [allowed as
> >>  normal], they chose (a).
> >
> >I think the characterization of the problem as being about optional
> >characters is too narrow.
> 
> No, that's exactly the problem we are discussing now. You might have 
> other problems you want to discuss, of course.

okay, we have different ideas about the nature of the discussion at
hand.  (though the subject line *does* say optional characters...)

it's of course fine to discuss optional characters but I suspect we 
need more than that to solve the transcription problem.

> >   The general problem is one of ambiguous
> >spellings resulting from transcription.  Optional characters are only
> >one example of this kind of ambiguity.
> 
> Correct, and it is the one we are discussing now. There is general 
> agreement that common misspellings or mis-identification of letters 
> is simply too hard to deal with and therefore out of scope. 

hmm.  I don't see how IDNs can serve people's needs if people can't 
transcribe them with reasonable accuracy.  transcription is fundamental
to the utility of DNS names, so I assume it's also fundamental for IDNs.

I don't expect IDNs to take care of homonyms, alternate spellings,
or misspellings of a word.   typing in colour or culor should not 
match the name color.  but someone who reads language X and has a 
suitable keyboard should be able to transcribe an IDN written 
in language X and get the right results.  someone who reads and
speaks language X should be able to read aloud an IDN written in 
language X to someone else who reads and speaks language X, and that 
person should be able to write it on paper or type it in accurately.
(even if the IDN has to be spoken one letter/symbol/whatever at a time)
a service that wants to make itself accessible by folks who read
language X should be able to choose an IDN that can be remembered
by readers of language X, and which they can write down, or type in,
from memory at some later time.

otherwise, what's the point of this exercise?  

we probably cannot eliminate all sources of transcription errors.
(lust like we allow both O and 0, and 1 and l, in DNS names) but there 
are other sources of transcription errors than just optional 
characters.  I think we need to know more about the nature of the 
problem in various languages before we declare them "too hard to deal 
with."

> >I'm not sure whether Unicode Consortium is the correct or only source
> >that should be consulted, but no matter who is consulted, the question
> >shouldn't be expressed solely in terms of optional characters.
> 
> Could you recommend who else you think is a good source?

no, I'm not personally familar enough with the organizations to be
able to recommend any of them.  

Keith