Re: [webfinger] Webfinger and URI vs IRI

["Martin J. Dürst" <[e-mail]>]

Hello everybody,

On 2013/07/23 6:27, Paul E. Jones wrote:
> Barry,

> The reason I raise this is that RFC 5988 refers to the target IRI (the
> “href” in WebFinger link relation) and context IRI (the “subject” and
> “aliases” in WebFinger).  Only ASCII is used in some protocols, so the
> IRIs must be formatted as URIs.

> However, JRD is JSON and, therefore, Unicode.  Thus, we could easily
> accommodate links like this:

>    {
>
>      "rel" : "test2",
>
>      "href" : "http://example.org/私の 文書.txt"
>
> }

> As opposed this form:

>    {
>
>      "rel" : "test2",
>
>      "href" :
> "http://example.org/%E7%A7%81%E3%81%AE%20%E6%96%87%E6%9B%B8.txt"
>
> }

> I have no strong preference, but the text did have IRI mentioned in one
> place in the JRD spec section, but it was not consistent through the
> document.  Everywhere else, we specified URI.

> So, if IRIs are truly only for presentation,

That's clearly not the case. IRIs are used in HTML and other places.

> then the latter example above
> should be what WF servers return.  The query target is always a
> percent-encoded URI, so it’s a non-issue.

For most of you, the differences between the above two examples are 
mostly irrelevant, and the second one may even look more familiar. But 
for those who can read the first one (Japanese, although the space is 
highly suspicious, because Japanese doesn't use spaces), the first one 
is very clear, whereas the second one is complete gibberish.

As a slightly related example, one could write
      "rel" : "test2"
as
      "rel" : "%74%65%73%74%32"
and it would provide about the same level of useless obscuration.

Please stop this "only for presentation" myth that essentially means 
that everything is legible as long as it's English.

Regards,   Martin.