Touchstones for "Mapping"

[Michel SUIGNARD]

From Mark Davis
> I think the issue is around what is meant by "storage". Look at a snippet of text:
>
> href="http://εύβοια.el"
>
> I just stored both of those URLs. Where? Well, in my GMail storage,
> in the idna message archive, on my machine (in Gears), on any recipient's machine
> (until deleted), eventually in the search engine caches, and the wayback machine.

I think both Vint and you have a point.

For all practical purpose, your text snippet is an IRI. IRI are so far only presentation elements for underlying protocol elements through a well defined mapping (see RFC 3987). Any application that use IRI in storage are supposed to map domain names detected in an IRI string (and that's not always easy) into a DNS complying name before processing it for DNS purpose.

The problem is that IRI is still stuck with IDNA-2003 definition. Until this is fixed, we are in a grey area, and all apps that have to convert IRI to URI have an issue. In that aspect A-labels are safer, unless applications are careful at storing U-label that are deemed to be identical with either IDN-2003 or IDNA-2003. Hopefully, implementers will be careful, otherwise we could be in a world of hurt.

Michel