- From: C. M. Sperberg-McQueen <cmsmcq@acm.org>
- Date: Fri, 01 Aug 2003 20:54:59 -0600
- To: public-qt-comments@w3.org
- Cc: W3C XML Schema IG <w3c-xml-schema-ig@w3.org>
Dear colleagues:
The XML Schema Working Group congratulates the XML Query and XSL
Working Groups on their progress, and in particular on the Last Call
draft of "XQuery 1.0 and XPath 2.0 Functions and Operators".
We have not been able to review the last call draft in as much detail
as we would have liked, but for what they are worth our comments are at
http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html (an
ASCII version is reproduced below for the convenience of those with
access to their email but not to the Web).
We apologize for the tardy arrival of these notes.
-C. M. Sperberg-McQueen, for the W3C XML Schema WG
................................................................
[1]W3C [2]Architecture Domain [3]XML | [4]XML Schema | [5]Member
Events | [6]Member-Confidential!
W3C XML Schema WG
Notes on XQuery 1.0 and XPath 2.0 Functions and Operators
1 August 2003
_________________________________________________________________
* 1. [7]Schema-related issues
+ 1.1. [8]Alignment of date/time values
+ 1.2. [9]The type anyAtomicType
+ 1.3. [10]The type untypedAtomic
+ 1.4. [11]Alignment on strings and URIs
+ 1.5. [12]Whitespace handling and lexical forms
+ 1.6. [13]Negative zero
+ 1.7. [14]Totally ordered Booleans
* 2. [15]Other technical issues
+ 2.1. [16]The fn:base-uri property
+ 2.2. [17]Alignment of references
+ 2.3. [18]Characters and collation units
+ 2.4. [19]Surrogate pairs and Unicode scalar values
+ 2.5. [20]Definition of whitespace
+ 2.6. [21]Required normalization functionality
+ 2.7. [22]Case folding
+ 2.8. [23]Escaping URIs
+ 2.9. [24]The binary types
+ 2.10. [25]Minor items
o 2.10.1. [26]User control of collations
o 2.10.2. [27]Section 7.3.1.1 Examples
* 3. [28]Editorial notes
_________________________________________________________________
This document contains comments on the Last Call draft of 2 May 2003
of XQuery 1.0 and XPath 2.0 Functions and Operators transmitted to the
XML Query and XSL Working Groups on behalf of the XML Schema Working
Group. These draft comments have not been reviewed by the XML Schema
Working Group and do not necessarily command consensus within the
group; because we will not meet again until 28 August, the Working
Group directed at its meeting today that these notes should be
transmitted to the XML Query and XSL Working Groups without awaiting
review.
In addition to the comments below, please note that several of the
[29]general comments sent on 14 July relate to the functions and
operators and data model specifications. Some of those comments sent
earlier overlap with some comments below.
1. Schema-related issues
The comments in this section relate to the use of XML Schema in the
F/O specification and thus to the particular area of responsibility
borne by the XML Schema WG.
1.1. Alignment of date/time values
The provision for preserving timezone information in the values of
xs:dateTime, xs:date, and xs:time continues to concern us. We believe
that a discrepancy of this kind between F/O and XML Schema will hurt
users and impede uptake of both specifications.
We believe F/O and XML Schema need to align on this, either by F/O
changing to the XML Schema value space, or by changing the value space
as part of XML Schema 1.1, or by some other mutually agreed upon
solution.
1.2. The type anyAtomicType
We reiterate our concern over the introduction of anyAtomicType into
the type hierarchy. We believe that a discrepancy of this kind between
F/O and XML Schema will hurt users and impede uptake of both
specifications.
We believe F/O and XML Schema need to align on this, either by F/O
aligning with XML Schema 1.0 or by XML Schema 1.1 aligning with F/O.
1.3. The type untypedAtomic
We reiterate our concern over the introduction of untypedAtomic into
the type hierarchy. As with the other discrepancies, we believe
alignment of the QT specs and XML Schema is critically important.
Section 1.3.2 says xdt:untypedAtomic is used wherever the PSVI has
xs:anySimpleType; please note that in the PSVI, this will be the case
* when the element or attribute in question was declared as having
type anySimpleType
* when the attribute in question had no declaration and the schema
processor assumed the simple urtype for it in the course of lax
validation or error recovery
Note that elements will not be assigned the anySimpleType as their
type property in the course of lax validation or error recovery; they
will have xs:anyType instead. Your use of xdt:untypedAtomic for
xs:anySimpleType but not for elements which (a) lack child elements
and (b) are assigned to xs:anyType may lead to results which puzzle
some of your users; we believe you may wish to consider changing your
mapping rules to assign xsd:untypedAtomic to such elements.
1.4. Alignment on strings and URIs
The table at the beginning of section 2, Accessors, shows functions
which are intended (judging by their names) to return URIs and which
return values of type xs:string instead of xs:anyURI. Similarly,
various functions which accept URIs as arguments are given signatures
using xs:string as the type, which in turn necessitates ad hoc rules
of the form "If $collationLiteral is not in the lexical space of
xs:anyURI, an error is raised".
As you know from our inquiry to you in mid-July, it has been suggested
that in XML Schema 1.1 the xs:anyURI type be made a restriction of
xs:string. But for now, there appears to be a discrepancy between the
use of strings to represent URIs here and the provision of a distinct
(and, for typing purposes, disjoint) type in XML Schema 1.0.
We need to align on this.
1.5. Whitespace handling and lexical forms
In section 5.1, paragraph 4 reads in part: "If the argument to a
constructor function is a string literal, the literal must be a valid
lexical form for its type ... Whitespace normalization is applied
before validation ..."
In all the cases which immediately come to mind, if the argument is a
valid lexical form for a type, there is no need to perform any
whitespace normalization on it. In XML Schema, it is the result of
whitespace normalization, not the input to it, which must be a legal
lexical form; we believe readers will be less confused if your usage
of the terms and ours is consistent.
A possible rewording: "If the argument to a constructor function is a
string literal, then whitespace normalization is applied as indicated
by the whitespace facet for the datatype. The whitespace-normalized
string must be a valid lexical form for the type, as specified ..."
1.6. Negative zero
In section 6, a note explains that the value space of xs:float and
xs:double has been extended vis-à-vis that given by XML Schema, to
include a negative zero. The note also explains that the negative zero
will "never be obtained from the typed value of a node."
We believe this discrepancy is untenable, and we are not clear why it
has proven necessary to introduce it.
As far as we can tell by examining the specification, the spec
mentions different treatment for positive and negative zero only for
the functions described in section 6.4 (fn:floor, fn:ceiling,
fn:round, and fn:round-half-to-even): in the description of each of
these functions it is noted that if a zero is given to the function as
an argument, the sign of the zero returned as the value of the
function is the same as the sign of the zero passed in as an argument.
(The discussion of fn:ceiling mentions other cases when negative zero
is returned; the discussion of fn:floor passes over the analogous
cases in silence.) Other mentions of the signed zeroes in this
specification invariably specify either that something is true both
for positive and for negative zero or else that a constructor may
return either a positive or a negative zero.
Could you explain the motive for introducing this discrepancy with the
value space defined in XML Schema? Would it not suffice to observe
that IEEE 754 has both positive and negative zeroes, which are treated
as different machine representations of the same values in the
xs:float and xs:double value spaces, and (optionally) that the prose
occasionally mentions these distinct representations of zero in the
interests of alignment with IEEE 754, even though formally they are
the same value?
Is it essential to introduce an incompatibility with XML Schema here
instead of treating positive and negative zeroes as one value with two
machine representations?
1.7. Totally ordered Booleans
We do not believe that it makes sense to impose a user-visible
ordering on the Boolean data type. Can you explain the rationale?
This is a discrepancy between F/O and XML Schema which must, we
believe, be aligned.
2. Other technical issues
The comments in this section relate to technical issues other than the
use of XML Schema in the F/O specification; the XML Schema WG claims
no particular responsibility or expertise on these questions but
raises them because they seem to need attention.
2.1. The fn:base-uri property
In section 2.5, the first paragraph defines a base-uri property for
all node types: "Document, element and processing-instruction nodes
have a base-uri property.... The base-uri of all other node types is
the empty sequence."
The next paragraph begins by explaining what happens "If the accessor
is called on a node that does not have a base-uri property ..." If all
nodes have the property, how can such a node exist?
2.2. Alignment of references
XML Schema and the Functions and Operators spec should refer to the
same version of Unicode. At the moment, this appears not to be true.
2.3. Characters and collation units
The discussion of collation units in the second note of section 7.3
says that collation decomposes a string "into a sequence of units,
each unit consisting of one or more characters", and that various
comparison operations are performed on these units. The functions
fn:starts-with, fn:ends-with, fn:substring-before, and
fn:substring-after are all mentioned as operating on such a segmented
string.
The list of functions at the beginning of section 7.4, however,
describes them as operating on characters, not on the nameless
collation units consisting of one or more characters each. This looks
like a contradiction.
We believe that the general level of confusion is best minimized, and
the world becomes a better place, if in XML-related specifications the
word character is used always and only for the units of the Universal
Character Set defined by Unicode and by ISO 10646. The word should not
be used (however great the temptation becomes at times) to denote the
culturally specific units of writing systems (e.g. letters, symbols,
signs, graphemes, or what have you).
We suggest recasting the descriptions in 7.4 to describe the effect of
the functions in terms of the collation units, rather than in terms of
characters. In order to avoid repeating the phrase "the nameless units
of one or more characters into which a collation segments a string for
purposes of comparison", you may wish to define the term letter,
grapheme, collation unit, or thingy with that meaning.
2.4. Surrogate pairs and Unicode scalar values
Section 7.4.6 (like some others) has a note calling attention to the
fact that some implementations will represent characters with code
points higher than xFFFF by using surrogate pairs. You quite correctly
avoid using the term code point for the things which make up the
surrogate pair, since in section 7.1 you have defined code point as
excluding surrogates. But the term 16-bit values is not defined, as
far as we can tell.
Also, in Unicode 2 and 3 there are (as far as we have been able to
tell) no rules that forbid a double encoding of characters outside the
Basic Multilingual Plane (i.e. first representing them within the BMP
as surrogate pairs, and then encoding the sequence of BMP items in
UTF-8). Even if it is discouraged (and it is indeed outlawed in
Unicode 4.0), surrogate pairs might well show up not only in UTF-16
but also in UTF-8, where they will presumably be presented by
Unicode-oblivious character libraries not as pairs of 16-bit values
but as four-octet sequences whose intepretation in terms of Unicode
scalar values requires slightly special rules.
Note that the definition of code points given in section 7.1 agrees
with the definition of Unicode scalar values in Unicode 4.0 in
excluding the surrogate range, but not with Unicode 2.0 (the version
cited in your normative references), or Unicode 3, which define a
Unicode scalar value as "a number N from 0 to 10FFF[16]", without
leaving any gap for the surrogates.
2.5. Definition of whitespace
Section 7.4.10 defines the function fn:normalize-space as doing
various things to whitespace, but it does not define the term
whitespace. It should, since various definitions are possible.
The Unicode character database, for example, lists the following
Unicode characters as whitespace in the file PropList-3_1_0.txt:
* 0009..000D ; White_space # Cc [5] <control>..<control>
* 0020 ; White_space # Zs SPACE
* 0085 ; White_space # Cc <control>
* 00A0 ; White_space # Zs NO-BREAK SPACE
* 1680 ; White_space # Zs OGHAM SPACE MARK
* 2000..200A ; White_space # Zs [11] EN QUAD..HAIR SPACE
* 2028 ; White_space # Zl LINE SEPARATOR
* 2029 ; White_space # Zp PARAGRAPH SEPARATOR
* 202F ; White_space # Zs NARROW NO-BREAK SPACE
* 3000 ; White_space # Zs IDEOGRAPHIC SPACE
The XML specification defines a smaller set of characters as
whitespace, for purposes of whitespace normalization.
So some definition is definitely needed.
2.6. Required normalization functionality
Section 7.4.11 requires conforming implementations to support Unicode
normalization form NFC.
Why is normalization form W3C not also required?
2.7. Case folding
Sections 7.4.12 and 7.4.13 define functions for case folding.
Since case folding is not consistent across languages and locales, we
have grave doubts about the wisdom of this inclusion, and some members
of the WG would advise you to drop these functions, which are not and
cannot be language- and culture-neutral.
There is precedent: the decision to drop case-folding of names from
the design of XML resulted from the realization that every
case-folding algorithm available, including the use of the Unicode
case mapping tables, has an inherent cultural bias. The inclusion of
culturally and linguistically biased functions does not contribute to
achieving the goal of universal accessibility for the Web. Some
members of the XML Schema WG believe your spec should not go forward
with these functions in it.
If you retain these functions, you should at the very least warn users
that
* Results may violate user expectations (in Québec, for example, the
standard uppercase equivalent of "é" is "É", while in metropolitan
France it is more commonly "E"; only one of these is supported by
the function as defined).
* Many characters of class Ll lack uppercase equivalents in the
Unicode case mapping tables (we stopped counting at 150 or so);
many characters of class Lu lack lowercase equivalents.
* The two functions are not inverses of each other, so that for a
string S of upper-case characters, fn:upper-case(fn:lower-case(S))
is not guaranteed to return S, nor is
fn:lower-case(fn:upper-case(S)) for a string S of lower-case
characters. Latin small letter dotless i (as used in Turkish) is
perhaps the most prominent lower-case letter which will not
round-trip, as Latin capital letter i with dot above is the most
prominent upper-case letter which will not round trip; there are
others.
You may also wish to make the case mapping depend on the default or a
user-specified collation.
2.8. Escaping URIs
The rules for escaping URIs should be aligned across all W3C
specifications; otherwise, we will drive our users crazy.
We think that means that you should reference and implement the
algorithm specified in the XML Linking specification
([30]http://www.w3.org/TR/2001/REC-xlink-20010627/#link-locators) and
referenced by XML Schema, or the algorithm given in the W3C Character
Model specification (which was the same algorithm the last time we
looked).
In particular, some members of the XML Schema WG were surprised to see
that your algorithm escapes the percent sign in some cases but not
others; this does not seem to be a feature of the algorithm given by
XML Linking and by the Character Model.
That said, we believe that you do your readers a good service by
listing explicitly the affected characters. By suggesting that you
refer to the Linking/CharMod algorithm, we do not mean to suggest that
you should make your spec less useful by omitting these lists.
(Editorial note: it would perhaps be useful to some readers to have a
brief discussion of why the advice given in the last paragraph should
be followed; our readers did not understand the rationale for this
advice.)
2.9. The binary types
Section 12.1.1 says that op:hexBinary-equal returns true if its
arguments "are of the same length and contain the same code-points";
similarly in 12.1.2 for op:base64Binary-equal.
The term code-point was defined in section 7.1 as denoting integers
between 0 and 1114111 (x10FFFF), with a gap in the range where Unicode
surrogates occur. It seems to be used here to denote what other
specifications refer to as octets (bit strings of length 8).
Taking the term code point in the sense of `octet', the definition
still does not match our intuitions of what an equality test on binary
data must do: it is not enough that each argument contain the same
octets; they must contain them in the same order.
Suggested rewording: "are identical strings of octets". If you wish to
avoid the word octet, "are identical bit strings" might do, although
it omits the relatively important fact that the values in question
must have 8×n bits for some integer n.
2.10. Minor items
2.10.1. User control of collations
Section 7.3 says in part "This specification does not use xml:lang to
identify the default collation, in part because collations should be
determined by the user of the data, not (normally) the data itself,
and because ..."
The second reason given is sound. The first (collations should not
normally be determined by the data) is often advanced as a principle,
but does not seem to all members of the XML Schema WG to be
universally true. We are thus grateful for the "(normally)" in the
sentence. But in any case, the first reason given here leads to a
non-sequitur: it would be a reason not to make xml:lang determine the
collation sequence without possibility of user override. But it does
not, even on its face, provide a reason not to use xml:lang to
identify the default collation. We suggest dropping the first reason;
the second suffices.
2.10.2. Section 7.3.1.1 Examples
The fourth example in section 7.3.1.1 says that
fn:compare('Strassen', 'Straße')
"returns 1 if and only if the default collation includes provisions
that equate `ss' and the (German) character `ß' (`sharp-s')." Unless
we have misunderstood the definition of the function, the return value
should also be 1 if the default collation sorts "ß" (sharp s) before
"s". Deleting the phrase "and only if" would remove the error.
3. Editorial notes
In the course of our work, some editorial points were noted; we list
them here for the use of the editors. We do not particularly expect
formal responses on these comments.
1. Definition of must. Section 1.1 defines must thus:
Conforming documents and processors are required to behave as
described; otherwise, they are non-conformant or in error.
Is the "or" inclusive or exclusive, or is "in error" intended as a
synonym or approximate synonym for "non-conformant"? Possible
alternatives: "otherwise, they are non-conformant and in error",
"otherwise, they are either non-conformant or else in error",
"otherwise, they are non-conformant, i.e. in error".
2. Definition of stable. In section 1.1, the definition of stable
says, inter alia:"Some other functions ... have an explicit
dependency on the dynamic context". Unless this means that they
accept an argument representing the dynamic context, it seems at
first glance as if explicit is here used with the meaning
`implicit'. Perhaps what is intended is that the documentation
will explicitly mention this dependency. Perhaps the best thing to
do would be just to drop the explicit; if you really wish to
stress the promise of documentation, perhaps read "Some other
functions ...have a depencency on the dynamic context ... These
functions are said to be contextual. [INS: Contextual functions
are always identified as such in their descriptions. :INS] "
3. The term back up. The phrase back up appears to be used several
times as a technical term (e.g. last paragraph of 1.7). What does
it mean?
4. The term QName. Some readers (including some members of the XML
Schema WG) are likely to find it disorienting for the term QName
to be used here as a synonym for expanded name or universal name,
and not with the same meaning QName has in the XML Namespaces
Recommendation. We recognize, however, that what is returned is
precisely a member of what XML Schema 1.0 defines as the value
space of the xs:QName type, so that the use of the term xs:QName
to denote (for example) the return type of the accessor
fn:node-name is not only unexceptionable but necessary for
consistency. We don't have a good solution for you here; we only
note the difficulty. Perhaps a note calling the reader's attention
to the issue would be in order (similar to the note on this topic
in the Data Model spec).
Some members of the WG suggest that this spec, like the Data
Model, should prefer the term expanded QName where possible, to
stress that what is referred to is the pair in the value space,
not the colonized Name in the lexical space.
5. No parameters and the empty list of parameters:
On first reading, the signatures of fn:string and fn:error suggest
an ambiguity to some readers: the call fn:error() appears to match
both the first and the second signatures.
Members of the WG who have studied XQuery more thoroughly assure
the rest of us that there is no ambiguity, so our purpose in
making this comment is merely to call your attention to an
editorial problem: it might be useful to explain to the reader why
the dual signatures showing no arguments and optional arguments
are not in fact ambiguous.
6. Section 2.3, first note: the word this seems to need an
antecedent; it is not clear to this reader, at least, what that
antecedent is. (It's also not clear what problem with blanks in
fragment identifiers is being adverted to.)
7. Raising errors: Section 3 para 1 reads in part: "The occurrence of
that phrase [sc. `an error is raised'] implicitly causes the
invocation of the fn:error function ..." This formulation seems to
involve a horrible clash of contexts: the phrase "an error is
raised" occurs in this document, and it occurs continuously from
the time of publication until the document ceases to exist (if
documents can ever cease to exist), while the error, one expects,
ought to be raised in a software system which implements the spec,
and should probably not be raised continuously from now until the
spec ceases to exist, if only because it would make it hard for
users to get work done. For the occurrence of a phrase in the spec
to cause the raising of an error in conforming software seems to
involve a rather unusual kind of action at a distance. To speak a
bit more seriously: perhaps the relevant part of the paragraph
could be recast, perhaps along these lines: "the phrase `an error
is raised' is used to describe the behavior of conforming
processors in certain situations. When such situations arise in a
running system, a conforming implementation of this specification
must invoke the fn:error function defined in this section." This
is not perfect, but we hope you get the idea.
8. Type promotion in multiple or single steps: Section 6.2 says "As
far as possible, the promotions should be done in a single step.
Specifically, when a decimal is promoted to a double, it must not
be converted to a float and then to a double, as this risks loss
of precision." [Emphasis added.] These two sentences appear to
contradict each other: is the rule about single-step conversions
required of conforming implementations ("must"), or recommended
without being required ("should")?
9. Code points: The note in section 7.1 identifies code points as
Unicode scalar values (which are in turn integers), but uses the
notation #x0000 and #x10FFFF to refer to the minimum and maximum
values. It's not terribly confusing in context, but strictly
speaking, this notation is defined in the XML specification as
denoting characters, not integers. I believe conventional
representations for hexadecimal numbers would write these values
as 0, 0H, x0, or 0x, and correspondingly 10FFFFH, x10FFFF, or
10FFFFx; there may be other hexadecimal representations you will
prefer. The Unicode specification writes 10FFFF[16].
10. v and w: Section 7.3 says "`uve' and `uwe' are considered
equivalent in some European languages"; this is unexpected. Are
you sure? Which languages?
11. Section 7.4 para 1: for "function" read "functions". Here and
elsewhere, we believe that sentences like "Several of these
functions use a collation" would do better if "a collation" were
replaced with a plural: "Several of these functions use a
collation." Unless, of course, all of these functions always use
the same collation.
12. Section 7.4.6.1, final example: forgive this observation if it's
clueless, but since there does not seem to be any addition
operator in the example (did we miss it?), it's not immediately
obvious what -INF + INF has to do with the interpretation of the
example.
13. Section 7.4.15, fn:string-pad: this seems an unfortunate choice of
names for a function which does not (despite its name) pad a
string with blanks or some other padding character(s), but which
simply replicates or copies the string multiple times. Could it be
renamed without excessive heartburn?
14. Section 7.4.16, fn:escape-uri: It would help minimize confusion if
the lists of characters which are or are not escaped gave the
character names as well as the characters themselves in quotation
marks. (In the paper copy used by one member of our review task
force, this bit of the spec was almost impossible to make out
without a magnifying glass.)
15. Section 7.5.3, fn:replace: The description of the function seemed
unclear:
The function returns the xs:string that is obtained by
replacing all non-overlapping substrings of $input that
match the given $pattern with an occurrence of the
$replacement string.
Replacing all occurrences of the pattern with an occurrence of the
replacement string seems to suggest an n for 1 exchange. For "all"
read "each". In the following paragraph, one occurrence of $input
is not marked as an identifier, one is.
References
1. http://www.w3.org/
2. http://www.w3.org/Architecture/
3. http://www.w3.org/XML/Group
4. http://www.w3.org/XML/Group/Schemas
5. http://www.w3.org/Member/Eventscal.html
6. http://www.w3.org/Member/#confidential
7. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e65
8. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e70
9. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e86
10. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e98
11. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e145
12. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e178
13. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e191
14. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e238
15. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e246
16. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e251
17. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e278
18. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e283
19. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e327
20. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e350
21. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e392
22. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e399
23. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e444
24. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e463
25. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e513
26. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e516
27. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e544
28. http://www.w3.org/XML/Group/2003/07/xmlschema-fo-comments.html#d0e577
29. http://www.w3.org/XML/Group/2003/07/xmlschema-query-notes.html
30. http://www.w3.org/TR/2001/REC-xlink-20010627/#link-locators
Received on Friday, 1 August 2003 22:55:50 UTC