RE: essential test cases?

At 05:42 PM 2000-06-14 +0200, Julian Reschke wrote:
>Dan Conolly wrote:
>
>> > There is nothing special about that test case, it is just an example
>> > using a relative URU. So presumably anyone who ever argued for the
>> > literal interpretation (which isn't just me, no matter how much you
>> > wished it were) must have argued that both uses were in the same "bat"
>> > namepsace (and so the document authors should probably have used
>> > better namespace names, such as absolute URI)
>>
>> Well... I was trying to avoid presuming or infering, having
>> had little success with that lately.
>>
>> I now take it that you dispute that the two bats should
>> be treated as distinct by XSLT implementations.
>>
>> Julian Reschke and Simon St.Laurent also spoke up, but
>> I can't quite tell if they meant to say "I disagree
>> that XSLT implementations should treat the bats as
>> distinct."
>
>OK, to clarify:
>
>I think that XSLT should treat them as equal if they are literally equal.
>This is not what the XPath spec says, but it seems that all current
>implementations work this way and (many? most?) people are happy with this.
>So I would suggest to bring XPath in line with the Namespace rec.
>

In what sense do you feel that XSLT sees whether "they are the same or
different"?

It seems to me that users of XSLT should have the opportunity to select on
a class that they both match or to select on two classes that distinguish
the two, as they please.

If the InfoSet contains both the literal attribute and the BASE for the
context, can one do that?

I agree that the hardwired "absolutize" functionality in XPath should be
removed.  My rationale was too mystical for Jon Cowan but I stand by it.
The abolutizing transformation is purely textual, but it ain't _XML_
syntax.  It's URI stuff.  And because the BASE can be inherited from
non-syntactic application contexts, it isn't reallly syntactic at all.

XPath should give paths terminating at everything in the InfoSet.  Should
not keep you from addressing the literal attribute value.

XSLT applications should be able to group or distinguish these two cases
ad_lib.

Al

Received on Wednesday, 14 June 2000 15:16:31 UTC