Re: [idn] I-D ACTION:draft-ietf-idn-uri-00.txt

["James Seng/Personal" <James@Seng.cc>]

You said there is a length problem of ACE. I am saying that ACE could be
design with the almost the same length as UTF-8 so it is not a concern.

> James Seng/Personal writes:
> > Without going to the details, it is possible to design an ACE which
> > has almost the resultant length as UTF-8
>
> Without gigantic tables of common character sequences? Without using
> punctuation or control characters that people can't type?

Why would ACE have these problems?

> Without a
> noticeable risk of misinterpreting existing ASCII names?

With a prefix as most ACE proposed, I say it is a "small risk".

> If it's possible to do this, go ahead! In the meantime, I'm going to
> focus on the proposals that have actually been worked out.

A lot of proposals have been proven to work to certain extend under certain
constraint. This includes ACE and UTF-8 and other even more horrible ones like
legacy encodings which is better not discussed etc. I have done all of them
and I still there is a better solution out there (e.g. John Klesin's I-D).

So the issue here is not "which works?", but "which works best?" and "what
constraint?".

> > if "10-byte limit" is your concern.
>
> There's nothing special about 10. My examples show that every extra byte
> poses a risk; we don't have the 63-byte safe harbor that some people
> have imagined. A user's 5-byte or 10-byte or 15-byte domain name can
> indirectly trip his software's 64-byte local-part length limit.

You are missing the point. See my first sentence.

-James Seng