[OpenID] Canonical OpenID url form

[Martin Atkins]

Peter Williams wrote:
> So the short form of the story is: use xri for unicode (and then transform the xri into an https hxri).
> 
> Its been a month since I studied xri (and thus have forgotten 80 percent of it). I recall there was a syntax to identify the address of the initial resolver. Is there a way tha this became the domain name componnt of the hxri
> 

XRI is not required to use non-ASCII characters in your OpenID Identifier.

What I take from this discussion is that the canonical form is 
percent-encoded UTF-8, but when it comes to displaying identifiers to 
end-users it can be transformed back into the real unicode characters 
using the same rules as browsers use.

It'd probably be a good idea to test how the various RP and OP libraries 
deal with this, though. I expect that in practice some implementations 
will get this wrong. I just tested the Net::OpenID::Consumer perl 
library and it only gets this right if the caller happens to pass it an 
already UTF-8-encoded string.