Re: same-document references

On 7/2/11 7:50 PM, Bjoern Hoehrmann wrote:
> As with any other interface you need to be clear in how you use it. I'm
> not even sure what the problem here really is, it seems common sense to
> me that you resolve internal references internally without new external
> retrievals

The problem is the definition of "internal".

> and even if you don't, RFC 3986 does not stop you if you do
> understand the implications and have good reasons.

Then, again, what's the point of this part of the spec?  It seems to be 
imposing some sort of normative requirements, but every time I try to 
figure out what those are, I'm told the section just doesn't apply....

> I would have thought that the problem would actually be the definion of
> what is a same-document reference to begin with

Yes, indeed.  That is the problem.

>
> One problem there is that if you have http://example.org/ with
>
>    ...
>    <thing id='example' />
>    ...
>    <context base='http://example.com/'>
>      <ref href='#example' />
>    </context>
>
> You might now say all of (under various, possibly wrong, assumptions)
>
>    *<thing>  is at http://example.org/#example
>    *<ref>  refers to http://example.com/#example
>    *<ref>  is a same-document reference

Indeed.  That is precisely the issue.

> Which some might find contradictory. The second statement is the most
> interesting if you think about how you would rewrite the document so
> all the base references are removed. But these issue arise long before
> "When a same-document reference is dereferenced for a retrieval action".

But they're irrelevant if all you do is define what a "same-document 
reference" is without then saying something about why those are 
special...  I don't care if the term is defined but then various other 
specs define how they want "same-document references" handled, I guess; 
they can always just define some other sort of processing.  But the 
combination of defining the term and imposing a requirement that seems 
to not match the implementations I know of is what worries me.

> In any case, this would be better discussed on the uri@w3.org list as
> changes to RFC 3986 are out of scope of the IRI Working Group

OK.  Again, I was responding to a general question about what parts of 
RFC 3986 currently cause compat issues....

-Boris

Received on Sunday, 3 July 2011 04:06:59 UTC