Re: A proposed solution

At 18:59 2000 06 08 -0400, Jonathan Borden wrote:
>Paul Grosso wrote:
>> At 17:50 2000 06 08 -0400, Jonathan Borden wrote:
>> >How about (requiring this condition)
>> >
>> >    * Existing documents which have a defined infoset will continue to
>have
>> >a defined infoset.
>
>    More variations on the same old problem....
>
>the document located at:
>
>    file://example.xml
>
>    <example xmlns:a="file://XXX/example.xml/YYY" xmlns:b="./YYY">
>        <zzzz a:a="1" b:a="2" />
>    </example>
>
>    is moved to:
>
>    file://XXX/example.xml
>
>    now it doesn't have an infoset (on the basis of not being namespace
>conformant)

Your example doesn't mention--and has nothing to do with--XML Base.

>> >
>> >... in order to meet this I think XBase will need to be scrapped.
>>
>> How so?  Not only doesn't XML Base have much anything to do
>> with the current discussion (see my previous posting [1]),
>> but existing documents couldn't have xml:base in them, so
>> how can existing documents be affected by XML Base at all?
>
>But 'new' documents produced by existing tools which create relative URI
>references might cause new problems.
>
>the document:
>
>    <example xml:base="file://XXX/example.xml"
>xmlns:a="file://XXX/example.xml/YYY" xmlns:b="./YYY">
>        <zzzz a:a="1" b:a="2" />
>    </example>

[Actually, the example doesn't show what you want; ./YYY with 
base of file://XXX/example.xml absolutizes to file://XXX/YYY.
Reread RFC 2396.  But I know what you mean.]

>has an infoset if the parser isn't XBase conformant, and not if the parser
>conforms to XBase....

You have the exact same situation if you don't use xml:base.  With
the absolutize option, you can set things up so that a document 
with relative URIs for namespace names will violate an attribute 
uniqueness constraint only when the document's base URI is a
certain set of values.  Nothing specifically to do with XML Base.

>I'm not actually suggesting XBase be scrapped, rather pointing out problems
>with XBase that are introduced *if* absolutization is specified. Such
>problems would not exist if either a) literal comparison  or b) forbid is
>specified.

XML Base is used to specify a base URI that is used during
the process of absolutizing relative URIs, and we know that
relative URIs are (1) useful and (2) need to get absolutized
in general (e.g., when they appear in links to other resources),
so it is completely out of line to malign XML Base here.

Your issue is whether namespace names (1) should be restricted
to non-relative ones (forbid) or (2) treated as non-URI-ish
strings (literal) or (3) absolutized.  If you don't absolutize,
a base URI is irrelevant, and if you do absolutize, you use
the base URI.  XML Base contributes to the calculation of
the base URI, not to the decision about whether you absolutize
namespace names or not.

Received on Thursday, 8 June 2000 19:43:01 UTC