Re: ZWNJ (was: Re: Just send UTF-8 with nameprep (was: RE: [idn]Reality Check))

[Roozbeh Pournader <roozbeh@sharif.edu>]

On Thu, 26 Jul 2001, Martin Duerst wrote:

> As far as I remember, currently ZNWJ is dropped by nameprep without
> complaining. I think that if that leads to problems as you describe
> above, we should very carefully reexamine this decision. I have
> already told the nameprep design team that I think that this should be
> moved to 'disallowed' to avoid surprises like the above. But I'm not
> sure moving it to 'disallowed' would solve all problems (assuming that
> the hyphen is used for concatenating words).

I think dropping it without complain should be much better than making it
disallowed. I really like to be able to display my Arabic domain name with
proper joining. Example: I register an Arabic equivalent of
RoozbehPournader.com, which will be:

	Reh Waw Zain Beh Heh Peh Waw Reh Noon Alef Dal Reh .something

This will create problems when displayed: Heh and Peh, while belonging to
different words, will join, which is a bad thing.

With the current nameprep rules (good), I can put a ZWNJ between Heh and
Peh, so they get separated, and nameprep will rip it out when resolving
the name, which is also fine.

Other solutions:

1. Disallow ZWNJ: bad. I should then use a hyphen between Roozbeh and
Pournader, which I don't like.

2. Allow ZWNJ in names: worse. Someone will put a ZWNJ between two letters
that do not join (like Reh and Alef), leading to no difference in shape,
by a difference in resolving.

3. Rip ZWNJ in nameprep, but also have some DNS (or WHOIS) mechanism to
get to the non-nameprepped name: great. This way any protocol can get to
the original name if it wants.

I want to recommend the third option. How this may happen, I don't know.
But this can be considered a reason for having the reverse mechanism in
place, which was the subject of the thread I replied to.

roozbeh