W3C home > Mailing lists > Public > public-qt-comments@w3.org > November 2005

XQTS: role-enum's console-log & output-file/@role

From: Frans Englich <frans.englich@telia.com>
Date: Sun, 13 Nov 2005 08:51:24 +0000
To: public-qt-comments@w3.org
Message-Id: <200511130851.24685.frans.englich@telia.com>


Hello,

A couple of question regarding the role attribute & related:

How is console logging supposed to work, from the perspective of the schemata? 
From what I understand, XQTSCatalog.xsd instances somehow should specify 
files for console-logging.

The problem I see is that one would have to add output-file elements to 
XQTSCatalog.xml for each test case(with a role="console-log" attribute) in 
order to get it to work. And after that, one would have the problem that test 
results would be specified in XQTSCatalog.xsd instances, not XQTSResult.xsd.

That leads into a general thinking of XQTSCatalog.xsd which I fail to 
understand: that XQTSCatalog.xsd is a combination between specifying 
input(tests) and output. For example, the "output-file" element when carrying 
'role="principal"' has the meaning:

"This is an output (either text or XML) that will contain the query results. 
If the processor invocation sequence accepts a filename for results, this 
name may be passed, possibly prefixed by a partial directory path to allow 
storage of the results in a separate directory tree."

I think this runs into the same problem as the 'console-log' case. I neither 
see why it is needed. Storing the results is not necessarily a necessity for 
producing a valid Results.xsd instance. Some harnesses perhaps save the 
results and compare against the base line on the file level, but that's 
implementation dependent, and I don't see why XQTSCatalog.xsd must be 
involved in specifying that.

I think this double role of XQTSCatalog.xsd is unfortunate, because 
Results.xsd instances are not independent containers of the produced output.

In absense of deeper knowledge, I think this would make sense to me:

* The "console-log" enumerator of the role-enum type is removed, and instead 
is an optional "message-output" element added as child to the test-case 
element in XQTSResult.xsd, which contains console output(if any).

* If serializations of result trees are of interest to store, do the 
specification of that in Results.xsd. Either in the XML file as the proposed 
change for "console logging", or reference external files as currently done 
in XQTSCatalog.xsd.

* The above changes would render the role attribute of the output-file element 
unnecessary; "output-file" would simply specify the baseline, as it currently 
do. If intrusive changes wasn't an issue, I would rename the element.

Of course, my general lack of insight in the format is likely the explanation 
to the problems I see.


Cheers,

		Frans
Received on Sunday, 13 November 2005 08:40:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 27 March 2012 18:14:42 GMT