[I18nrp] draft-ietf-iri-comparison-02

[Larry Masinter <[e-mail]>]

https://tools.ietf.org/html/draft-ietf-iri-comparison-02
"Comparison, Equivalence and Canonicalization of Internationalized Resource
Identifiers"
Went through a few revisions as draft-masinter-iri-comparison too.

This draft was abandoned when the working group closed (for lack of
participation), but should be revived.

The IRI comparison/canonicalization problem subsumes the domain name and
email address problems, pretty much. That is, if you could define meaningful
way of determining equivalence of two IRIs, you could trivially use that to
define equivalence for domain names (prepend "http://") and for email
addresses (prepend "mailto:")  So why not take on the bigger problem?

Basically, any process or application that involves a human viewing a
rendering of a string of characters and comparing it to a rendering of
another string, or re-entering it,  or remembering it can depend on the user
sufficiently that it is not possible to define an equivalence or
canonicalization algorithm that would be uniformly acceptable.

My simple idea for CanIUse.name site would be to let a user input a string,
render the input in a couple of simple type styles, and use OCR with
language auto-detection; if you got back the same string, it would be a name
that's pretty safe.   All of the known problems (normalization, confusing of
I l 1 |, 0 O, seem to be handled pretty well.