W3C home > Mailing lists > Public > public-sml@w3.org > January 2009

new draft impl reports for today's telecon

From: John Arwe <johnarwe@us.ibm.com>
Date: Thu, 22 Jan 2009 10:31:15 -0500
To: public-sml@w3.org
Message-ID: <OFCD76C810.EAEEEF40-ON85257546.00533BEA-85257546.005541ED@us.ibm.com>
TestCaseImpl... generated from the xml contained in the zip Kumar emailed 
last night.
Formatting changes since last time:
1. added row number after test case name to make it easier to refer back 
to xlsx when needed
2. completely blank lines suppressed
3. counts of total, blank, non-blank lines added at end

Net: we have 145 tests in the current xlsx.  The XSLT flags 8 lines as in 
need of attention:

19: MS result is blank.  If we go with the "as test case is currently 
coded in cosmos cvs" approach to expected results, the expected result is 
INvalid, which should match MS's result given Kumar's remarks. I verified 
there is a bug in cosmos causing it to skip XML schema validation on this 
test (no target namespace + no aliases triggers it; since the majority of 
the tests have a target namespace, this should be a fairly isolated case).

29: Multiple sml:uri's in one reference, so if we are consistent with 
earlier decisions we will remove this row.

76, 77, 96-99 flagged because their primary feature is xmlbase and we have 
not fully resolved the optional features approach, although we have a 
proposal to discuss.  I did look through them, and (as currently coded in 
cvs) the expected results for an implementation -not- supporting xml:base 
should be (I think :-):

76: invalid (a relative SML reference exists and xml:base is the only 
mechanism used to specify the base URI, -and- the relative ref is in an 
element governed by sml:targetRequired=true)

77: valid (unresolved but not targetRequired)
96-99: valid (some same as 77, some same-doc refs, some have no relative 
ref anyway) 
For this latter list, 77+96-99, we should add remarks to suggest cosmos 
alter the tests so a non-supporting impl would find the model invalid.  As 
Kumar said, if they do so later we can always revise the impl test report 
and point to the revised testcases with different results.

I am also attaching the features report (unchanged, since there has been 
little discussion of the proposal for how to deal with optional features - 
hint hint).  If we achieve consensus on the "as currently coded" approach 
and report results for all impls against all test cases, we still need MS 
results on 6 rows (use the filter in the xlsx column header to see them, 
de-selecting T and F).

Best Regards, John

Street address: 2455 South Road, P328 Poughkeepsie, NY USA 12601
Voice: 1+845-435-9470      Fax: 1+845-432-9787

Received on Thursday, 22 January 2009 15:31:59 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:56:14 UTC