Test suite proposals

Here are some proposals for resolving issues: TS05,TS14,TS15,TS16,TS18

TS05

  Japanese test are intended to test encoding and japanese character.
  There is a declaration for <!ENTITY lt "<"> is totally not useful
  in the scope of the test intention.  And they  are already in the
  XMLTEST suite.

Resolution to remove the offending declaration.  Action to also
note that a future test case should be created to test that parsers
reject such declarations as not well-formed.  (BTW, what is the
process for us to track requests for future tests that are not
targeted for this next release of the suite?)

TS14

  Output file has PI before the DTD (wrong per SUN canonical form
  definition)-  Not yet updated.The only definition for the Sun
  canonical form that I have found is the one packaged with the
  conformance tests.  This is labelled "Draft 1" and does, as you
  state, indicate that the DTD should appear first.   The link to
  the online document is broken.  I wonder if all of the current
  implementations currently emit PIs in the order (relative to the
  DTD) that they appear (as ours does).

Propose to leave this as is and update cxml.html to conform to existing
practice.  As observed, implementations of the test harness emit the
DOCTYPE immediately before the root element, which would cause it to
follow any PIs before the root element.  There is nothing to be gained
at this point by declaring those implementations as not conforming to a
non-standard "specification" of a canonical form.

TS15

  <test names omitted> contain non-deterministic content models.

Proposed resolution to remove those content models is acceptable.

TS16

  Related to normalization of attribute values. The canonical output of
  the attribute "nmtokens" is wrong as it is missing a space (#x20)
  character.

  There has been some confusion about the validity of tokens separated
  by character references to whitespace (which depends on the definitions
  of Names and Nmtokens).  They were outlawed by first edition erratum 62,
  which was reversed by erratum 108 (apparently) because of a
  misunderstanding) and have been outlawed again by second editions erratum
  20.  The end result is that there examples are not valid and should
  therefore not have any canonical output!

Since these tests are supposed to be used for valid standalone document
tests, and not for finding errors with invalid character referenced in
attribute values, the character reference &#x20; should simply be removed
from "nmtokens" attribute.

TS18

  xmltest/not-wf/sa/176.xml
  xmltest/valid/sa/093.xml

  Input files changed line-ends.

Not sure what this is referring to, unless it is just another simple
problem with
EOL handling w.r.t. CVS.  If there is a "historical" version of these
files, say
from an earlier release of the suite, that was identical except for line
ends, I
would agree that we should not be arbrarily changing such files.

Regards,
Glenn

Received on Monday, 25 March 2002 16:20:40 UTC