Schema imports in MinimalConformance

[Xavier Franc]

Dear XQTS developers,

There are still many "import schema" declarations in MinimalConformance
tests (around 250, nearly 2%).

These tests will always fail with an implementation that does not
support Schema:
"If an XQuery implementation does not support the Schema Import Feature,
it MUST raise a static error [err:XQST0009] if it encounters a schema
import."

IMHO those tests should be moved into the "Optional" part, rather than
be adorned with an "expected-error" clause.

I am looking forward to seeing this done in the 1.0 version.

Best regards.

PS: I intentionally dont add a bug into bugzilla: I am not comfortable
with this tool, and I think it is up to the
XQTS developers to manage the bug database. Furthermore there are
already several reports concerning this topic.

[Frans Englich-2]

On Thursday 13 July 2006 22:50, Xavier Franc wrote:
> Dear XQTS developers,
>
> There are still many "import schema" declarations in MinimalConformance
> tests (around 250, nearly 2%).

Yes, it seems to be this one:

http://www.w3.org/Bugs/Public/show_bug.cgi?id=2666

> These tests will always fail with an implementation that does not
> support Schema:
> "If an XQuery implementation does not support the Schema Import Feature,
> it MUST raise a static error [err:XQST0009] if it encounters a schema
> import."
>
> IMHO those tests should be moved into the "Optional" part, rather than
> be adorned with an "expected-error" clause.
>
> I am looking forward to seeing this done in the 1.0 version.

Unfortunately I must say it's highly unlikely this will be fixed in 1.0. Yes,
the issue has been open for long, and I would also very much like to see it
fixed.

The reason is that we(the test suite task force) have problem getting in
contact with all our members. Vacations pokes holes in work flow this time of
year.

Hopefully, it will be fixed in a bug fix release that has been suggested to
follow the 1.0 release.

> PS: I intentionally dont add a bug into bugzilla: I am not comfortable
> with this tool, and I think it is up to the
> XQTS developers to manage the bug database. Furthermore there are
> already several reports concerning this topic.

You are of course very welcome to use this list. Many find using the bug
database practical. Let us know if there's any trouble you have so we can
help, if you like. But again, you're welcome to use this list.

(Personal response.)


Regards,

                Frans

[Wolfgang Hoschek-2]

What's the rationale for making a 1.0 release if it contains a long  
laundry list of known bugs, going back all the way to 0.8.0? (e.g.  
many unresolved whitespace issues, as of yesterday). 1.0 will be  
interpreted as a call for conformance, and I'd rather not conform to  
a buggy test suite. Why not call it 0.9.5? or 1.0-beta1?

Wolfgang.

On Jul 14, 2006, at 4:12 PM, Frans Englich wrote:

>
> On Thursday 13 July 2006 22:50, Xavier Franc wrote:
>> Dear XQTS developers,
>>
>> There are still many "import schema" declarations in  
>> MinimalConformance
>> tests (around 250, nearly 2%).
>
> Yes, it seems to be this one:
>
> http://www.w3.org/Bugs/Public/show_bug.cgi?id=2666
>
>> These tests will always fail with an implementation that does not
>> support Schema:
>> "If an XQuery implementation does not support the Schema Import  
>> Feature,
>> it MUST raise a static error [err:XQST0009] if it encounters a schema
>> import."
>>
>> IMHO those tests should be moved into the "Optional" part, rather  
>> than
>> be adorned with an "expected-error" clause.
>>
>> I am looking forward to seeing this done in the 1.0 version.
>
> Unfortunately I must say it's highly unlikely this will be fixed in  
> 1.0. Yes,
> the issue has been open for long, and I would also very much like  
> to see it
> fixed.
>
> The reason is that we(the test suite task force) have problem  
> getting in
> contact with all our members. Vacations pokes holes in work flow  
> this time of
> year.
>
> Hopefully, it will be fixed in a bug fix release that has been  
> suggested to
> follow the 1.0 release.
>
>> PS: I intentionally dont add a bug into bugzilla: I am not  
>> comfortable
>> with this tool, and I think it is up to the
>> XQTS developers to manage the bug database. Furthermore there are
>> already several reports concerning this topic.
>
> You are of course very welcome to use this list. Many find using  
> the bug
> database practical. Let us know if there's any trouble you have so  
> we can
> help, if you like. But again, you're welcome to use this list.
>
> (Personal response.)
>
>
> Regards,
>
> Frans
>

On Friday 14 July 2006 23:12, Wolfgang Hoschek wrote:
> What's the rationale for making a 1.0 release if it contains a long
> laundry list of known bugs, going back all the way to 0.8.0? (e.g.
> many unresolved whitespace issues, as of yesterday). 1.0 will be
> interpreted as a call for conformance, and I'd rather not conform to
> a buggy test suite. Why not call it 0.9.5? or 1.0-beta1?

Personally I think your comment has merit. Labeling it 0.9.5 is a whole
different thing compared to 1.0, as outlined.

I can't comment on this as whole(because I don't have insight in the full
scope, such as the impact of changing the road map and dead lines) so I'll
leave this to Andrew for replying.

Thanks for asking "why?"!

(Personal response.)


                Frans

[David Carlisle]

[Note I'm not in the Working Group, just a fellow Test suite user]

> There are still many "import schema" declarations in MinimalConformance
> tests (around 250, nearly 2%).
>

In addition to that there are many more which don't explictly import a
schema but have hidden dependency on schema support as the expected
result files assume that the input tree has type annotations from schema
validation.

I've argued often that all these schema dependencies should be removed.

> These tests will always fail with an implementation that does not
> support Schema:

This isn't strictly true as the test suite guidelines allow you to
change the schema import to a namespace declaration before running the
test. So it is possible to pass some of those tests.
http://www.w3.org/XML/Query/test-suite/Guidelines%20for%20Running%20the%20XML%20Query%20Test%20Suite.html

>    This declaration can be customized in one of the following three ways:
>
>    1. Unchanged: use schema import as indicated in the original query
>    2. Remove the schema import declaration from the query, and add
>    namespace declaration using same name and URI to statically known
>    namespaces before the query is executed.
>    3. Replace schema import with namespace declaration using same name
>    and URI.



This licence to change the query text is not particularly helpful though,
and I strongly urge that it be _removed_ from the guidelines.

See bug
http://www.w3.org/Bugs/Public/show_bug.cgi?id=2666
for the most recent of several bug reports in this area.

If changing the schema import to a namespace declaration allows the test
to run correctly then the schema dependency is spurious, so the test
should be changed as distributed just to use a namespace declaration.

If changing the schema import to a namespace declaration causes the
test to fail (as it requires type or element declarations) then
suggesting that implementors make this change before running the test is
just misleading and unhelpful. The test needs to be moved to the
optional schema import section of the test suite.


David

[Wolfgang Hoschek-2]

On Jul 15, 2006, at 2:46 PM, David Carlisle wrote:

>
>
> [Note I'm not in the Working Group, just a fellow Test suite user]
>
>> There are still many "import schema" declarations in  
>> MinimalConformance
>> tests (around 250, nearly 2%).
>>
>
> In addition to that there are many more which don't explictly import a
> schema but have hidden dependency on schema support as the expected
> result files assume that the input tree has type annotations from  
> schema
> validation.
>
> I've argued often that all these schema dependencies should be  
> removed.
>
>> These tests will always fail with an implementation that does not
>> support Schema:
>
> This isn't strictly true as the test suite guidelines allow you to
> change the schema import to a namespace declaration before running the
> test. So it is possible to pass some of those tests.
> http://www.w3.org/XML/Query/test-suite/Guidelines%20for%20Running% 
> 20the%20XML%20Query%20Test%20Suite.html
>
>>    This declaration can be customized in one of the following  
>> three ways:
>>
>>    1. Unchanged: use schema import as indicated in the original query
>>    2. Remove the schema import declaration from the query, and add
>>    namespace declaration using same name and URI to statically known
>>    namespaces before the query is executed.
>>    3. Replace schema import with namespace declaration using same  
>> name
>>    and URI.
>
>
>
> This licence to change the query text is not particularly helpful  
> though,
> and I strongly urge that it be _removed_ from the guidelines.
>
> See bug
> http://www.w3.org/Bugs/Public/show_bug.cgi?id=2666
> for the most recent of several bug reports in this area.
>
> If changing the schema import to a namespace declaration allows the  
> test
> to run correctly then the schema dependency is spurious, so the test
> should be changed as distributed just to use a namespace declaration.
>
> If changing the schema import to a namespace declaration causes the
> test to fail (as it requires type or element declarations) then
> suggesting that implementors make this change before running the  
> test is
> just misleading and unhelpful. The test needs to be moved to the
> optional schema import section of the test suite.
>
>
> David

Well said, thanks!
Wolfgang.