Re: IRIs everywhere (including XML namespaces)

On Thursday, October 10, 2002, 11:16:09 PM, Jonathan wrote:


JM> I agree.  None of these have normative references to IRIs, but try to
JM> incorporate the primary concept that users do not perform %-escaping
JM> prior to inserting the IRI into an XML document, but that the processor
JM> applies %-escaping when needed to interface with a component requiring a
JM> strict URI.

JM> I also think your careful reading of XLink role and arcrole descriptions
JM> is correct; my mistake.  But it is my recollection that we tried to make
JM> role and arcrole consistent with href, which the wording does not
JM> express.


So, it would be an oversight (or perhaps, one of those cut and past
errors that would be mitigated in future by a normative IRI spec and
an architectural principle recommending its use) that role and arcrole
do not permit IRI currently?

JM>   I'll ask the XLink group to search their memories.  In the
JM> meantime, it would be useful to hear the TAG opinion on whether using
JM> URIs instead of IRIs in role and arcrole attributes was a mistake.

I can't give you a TAG opinion, at this point, but from reading the
summary below I would say, in hindsight, that yes it was a mistake
because there was no particular advantage to be gained by having to
hexify them. So, in my personal view, serious consideration should be
given to making XLink role and arcrole take an IRI in any subsequent
revision.

>> -----Original Message-----

>> >- XLink role and arcrole attributes [5]
>> 
>> This is not accurate. The XLink spec is clear that illegal characters
>> must be escaped before their insertion in the role and arcrole
>> attributes:


-- 
 Chris                            mailto:chris@w3.org