values of Type

Drummond Reed drummond.reed at
Thu Dec 8 01:26:33 UTC 2005

The timing of this conversation was perfect because the XRI Resolution
editors just had a discussion about this today and arrived at the same
conclusion: the spec will say that all xrd:Service/xrd:Type values SHOULD be
absolute URIs (or IRIs or XRIs in URI normal form) for the very reasons
Martin mentions.

As I mentioned, we can't enforce this beyond this SHOULD language because
the W3C anyURI type explicitly allows relative as well as absolute URIs. But
at least all our example XRDS documents (that means you, Andy ;-) should use
only absolute URIs.


-----Original Message-----
From: yadis-bounces at [mailto:yadis-bounces at]
On Behalf Of Johannes Ernst
Sent: Wednesday, December 07, 2005 11:40 AM
To: Martin Atkins
Cc: yadis at
Subject: Re: values of Type

I agree with Martin.

On Dec 7, 2005, at 0:18, Martin Atkins wrote:

> Drummond Reed wrote:
>> However I fully agree that Type values SHOULD be absolute in order  
>> to reduce
>> the potential for collision. So Andy's examples SHOULD have
>> "+authentication" and "+photos", as these are absolute XRIs. (Note  
>> that
>> these are equivalent to "xri://+authentication" and "xri:// 
>> +photos", i.e.,
>> the "xri://" part is optional with absolute XRIs that start with  
>> global
>> context symbols (=, @, +, $, !. This was conscious design decision to
>> simplify human usability, e.g., to type a global XRI in an editor  
>> or address
>> bar only requires a single symbol character.)
> While this almost certainly isn't the place for this concern, I feel
> that such "shortcuts" should only apply in places where users type  
> in a
> value and it's used transiently.
> Browsers allow users to omit the "http://" part of an HTTP URL when
> entering one into the address bar, but it's still advisable to always
> *store* it as a full, canonical URL. Likewise, I think it's  
> advisable to
> require XRIs that are stored in some way, such as in these XRD
> documents, to be fully-qualified.
> The difference here is one of context. In a browser address bar, we  
> can
> guess that most entered URLs are going to be HTTP ones. Likewise in
> something that operates on XRIs it can assume that the thing being
> entered is probably an XRI. However, in a field marked "anyuri"  
> there is
> no such context, and processors shouldn't be forced to know the
> canonocalization rules for every single URL scheme out there in  
> order to
> make sense of a document.

Johannes Ernst

More information about the yadis mailing list