The Catalog of Vague, Draft 1

[David_Marston@lotus.com]

Here is my first attempt to list all the areas where the XSLT and XPath
Recommendations are not precise enough for the OASIS conformance committee
(or XSLT processor developers) to tell what is correct behavior. This
differs from the previously-published list of discretionary items because
those were explicitly granted by the specs. These are "discretionary" only
insofar as a developer has to use their own discretion in interpreting the
specs.

This list draws upon our experience at Lotus plus vagueness itemized in
Mike Kay's book. I expect that a few more items will emerge, and the XSL
Working Group may say that some of the items below are actually answered if
I would only read the spec more precisely. Scott or Eduardo, let me know if
you take this list to the WG. All: please send corrections and additions to
me.

Each item has a handy (?) mnemonic label. I use the term "NaN-string" as an
abbreviation for "any string that cannot be converted to a number," which
of course includes "NaN" itself.

[attrib-after-comment] When populating an element, should it be a required
error to invoke xsl:attribute after xsl:comment?
[attrib-after-PI] When populating an element, should it be a required error
to invoke xsl:attribute after xsl:processing-instruction?
[attrib-set-merge] Must the process of merging same-named attribute sets be
completed before the set is used anywhere, including in the definition of
other attribute sets? What if an xsl:include is situated between two
xsl:attribute-set elements?
[attrib-set-not-exist] Is an attempt to use-attribute-sets with a
non-existant set an error, ignoreable, or developer's choice between those
two?
[choose-unselected-when-error] The xsl:when cases in a choose must be
selected in document order, but a multi-threaded processor might evaluate
the when tests in parallel. Must the effects of all unselected when tests
be neutralized?
[excluded-prefix-needed] What should be the effect of an attempt to emit an
element or attribute bound to a namespace on the list of excluded
namespaces?
[fallback-top-level] Can xsl:fallback be implemented inside a top-level
"extension" element? If so, must it be detected and instantiated as with
instructions?
[for-each-variable] Can xsl:variable inside a for-each loop be set to a new
value on every iteration?
[func-document-self-include] Does document("") refer to the containing
stylesheet without included stylesheets?
[func-document-second-empty] If document() is called with two arguments and
the second is an empty node-set, should the second argument be ignored?
[func-system-property-namespace] What rules apply to use of either the
default or xsl namespaces on arguments to system-property (if the processor
developer wishes to add more properties)?
[include-position] Should xsl:include occurring after other top-level
elements (especially xsl:template, xsl:param, xsl:variable) confer all the
effects of its elements occurring later in the including stylesheet? If
not, which elements in the included document act as if they were in
positions in the including stylesheet other than the position of the
xsl:include?
[key-unique-name] Since "each key name may be thought of as distinguishing
a separate, independent space of identifiers." would multiple xsl:key
declarations with the same key name be allowed? If so, how do the multiple
declarations interact?
[message-full-template] The syntax of xsl:message shows content: template.
Does this include xsl:element, etc.?
[namespace-alias-intermediate] Should the namespace-alias be applied upon
first creation of the nodes? (Ref.: Kay, p. 235)
[namespace-alias-result-prefix] Should the result-prefix in
namespace-alias, if not #default, be the exact prefix used in the output?
[number-NaN] If xsl:number is called with a value set to positive or
negative infinity, or a NaN-string, how should it be formatted?
[number-negative] Should xsl:number render a negative number (specified in
value), even in alphabetic and Roman-numeral formats?
[number-zero] Should xsl:number emit a zero, even in alphabetic and
roman-numeral formats?
[output-text-newline] Should treatment of new-line characters be prescribed
for the text output method?
[pattern-document] Should the document() function be explicitly banned from
match patterns?
[sort-case-order] Does the case-order attribute of xsl:sort pertain to
ascending order, and its opposite to descending order?
[sort-NaN] Where should NaN-string values be placed in ascending and
descending numeric sorts?
[source-notations] What should the XSLT processor do about "notations" in
the source document? (Ref.: Kay, pp. 64-65)
[variable-top-level-full-template] When xsl:variable and xsl:param are used
as top-level elements, they can contain template instructions. Is there any
way in which these instructions are more restricted than the same
instructions in xsl:template? Specifically, is xsl:fallback available?
[with-param-full-template] The syntax of with-param shows content:
template. Does this include xsl:element, etc.?
[XPath-ceiling-NaN] Should ceiling() of any NaN-string return NaN?
[XPath-ceiling-negative-fraction] Should ceiling(-0.1) return positive or
negative zero?
[XPath-concat-one-node-set] If concat() is given one argument consisting of
a node-set, why not concatenate the string values of all the nodes? (Note:
some processors currently allow this.)
[XPath-contains-main-empty] If the first argument to contains() is empty
and the second argument is non-empty, should false be returned? Given that
every string contains the empty string, does the empty string contain
itself? (In other words, does contains("","") return true?)
[XPath-floor-NaN] Should floor() of any NaN-string return NaN?
[XPath-string-length-diacriticals] Should string-length() determine when
diacritical marks are not distinct characters?
[XPath-substring-diacriticals] Should substring() determine when
diacritical marks are not distinct characters?

Here's one that I cannot confirm:
[XPath-round-details] Mike Kay (p. 84) says that some behaviors of round()
are not specified. If so please be more exact.