W3C home > Mailing lists > Public > public-iri@w3.org > April 2003

Re: How to refer to IRI's

From: Ray Whitmer <raydwhitmer@aol.com>
Date: Thu, 24 Apr 2003 08:39:44 -0600
Message-ID: <3EA7F730.9040102@aol.com>
To: public-iri@w3.org

Thanks for your responses.

duerst@w3.org wrote:

> At 10:52 03/04/23 -0600, Ray Whitmer wrote:
> [...]
>> 1.  How do we refer normatively to the IRI specification now, since 
>> we are producing documents which are progressing rapidly.
> I think this is a W3C-specific question, which this public list
> may not be able to address completely. Various W3C specifications
> have done various things to describe IRIs. The purpose of this list
> is to work on and finish the IRI specification quickly, not to tell
> people what to do in the meantime.

Fair enough.  Unfortunately, I have not identified the correct list to 
ask about this.  This list seemed the closest, although that may have 
been an erroneous determination.  I18N lists do not seem correct, 
because these questions are not fundamentally internationalization 
questions -- the internationalization is already satisfied by IRIs and 
the question is how to replace URIs with IRI's or the next great thing. 
 URI lists did not seem correct, because IRIs and other possible 
follow-ons are not URIs.  Perhaps the tag list?


> In that case, the DOM just has to make
> sure that the values for the set/get allow characters outside
> US-ASCII, and can leave the interpretation of these values
> to the XML spec.) So some more discussion may be needed.

DOM has never specified that these values were necessarily checked, but 
the user of a specification has been told that these are URIs.  Just as 
a user who supplies a native filename argument would not expect a URI to 
work, even though it is possible to contemplate an implementation that 
allowed that, regularly supplying an IRI into an attribute or parameter 
which is called "URI" is quite problematic.  As we fix this, it seems 
short-sighted to take an argument or attribute that was previously named 
or described as a URI and now call the argument an IRI.  If we have to, 
I guess we make up our own generic name and description that will not 
break so easily again.


> One of the principles of IRIs is that they should not be used
> indiscriminately where URIs are used, but that a spec actually
> should explicitly say that a given parameter/attribute/... is
> an IRI. This does not match well with the idea of 'automatic update'.

But in practice, the idea of the automatic update (possibly with a few 
explicit exceptions) seems more likely.  This is quite consistent with 
not only with my own own inclination, but also with the feedback from 
the I18N working group on the DOM Core and Load and Save Modules:


C17) Substitute IRI for URI throughout.
LS5) Substitute IRI for URI throughout.

> Also, URIs in general are defined to be very flexible, and
> accommodate new functionality e.g. through new schemes. It is
> generally considered better not to impose strict syntax checks
> on interfaces that handle URIs to allow future evolution.

And we generally do exactly that where practical.  But, again, IRI's are 
not URI's.  The new URI schemes always followed the basic meta-rules, 
such as not permitting space and various other quite significant 
characters -- I am not worried about further evolution that is still a 
URI or IRI.  That is already covered.  I am worried about the next time 
new types are introduced.

> Also, I don't know of any other major issue than internationalization
> that is outstanding. But if there were any, it wouldn't be on
> this list that it would be discussed. So maybe asking on the
> uri@w3.org list will give you a better idea. 

The thing to discuss on the URI list might be extending the URI 
specification so that IRI's can be properly called URI's.  If this is 
not likely, it would not seem to me to be the proper forum.  The URI 
rules are still quite restrictive in a number of ways.  We suspect that 
there will be more out-of-the-URI-syntax-box enhancements.

Is discussion of IRI's considered a subset of the URI discussion, or 
outside of the URI discussion because IRI's are not URI's?  If it is 
outside, then it is not clear to me that that group discusses things 
that fall outside of their domain.  We are not asking about how to refer 
to URI's, but rather how to avoid referring to any specific RI scheme 
such as URIs or IRIs because of the situation this leaves us in.


Ray Whitmer
Received on Thursday, 24 April 2003 10:40:29 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:39:37 UTC