Re: Review of Manchester Syntax Document

Thanks for the review Mike.

> Below are my review comments for the Manchester Syntax document.  In
> addition to these comments, I made a single trivial spelling
> adjustment (diff at [1]).  Of the comments below, I believe (1) and
> (11) are most significant.

I've reordered the comments to keep the similar themes together and to
put the simplest changes first.

> 3) The BNF for floatingPointLiteral , exponent , decimalLiteral , and
> integerLiteral permit leading zeros.  Since nonNegativeInteger is
> already defined, leading zeros could be prevented.  It is not clear
> allowing them is beneficial.

Not changed, as leading zeros are allowed in other syntaxes, and are
benign. 

> 9) The entry for HasKey axioms in section 4.2 and the manchester form
> of the axiom do not reflect recent changes to the HasKey functional
> syntax (disambiguating object and data properties).

The Manchester syntax is not typed, so the form of the HasKey axiom will
not change (for this, but see below).

> 6) In the first bulleted condition of section 2.6 reference
> restrictions on the use of reserved vocabulary.  These restrictions
> apply to all OWL 2 ontologies, not just OWL 2 DL ontologies.

The first bullet was about the vocabulary for ontology and version IRIs,
which is DL-ish, not the presence of ontology and version IRIs, which is
OWL-ish, and guaranteed by the form of the Manchester syntax.  (However,
see below for major changes to this section.)

> 2) The description of whitespace in the grammer (section 2 para 4)
> permits but does not require whitespace between the quotedString and
> the '@..' of abbreviatedRDFTextLiterals.  While (I don't think) that
> this prevents parsing, I believe it was unintended (based on the
> restriction being present on typedLiteral) and that
> abbreviatedRDFTextLiteral should be added to the list of non-terminals
> in which whitespace is not allowed.

Fixed. Now no whitespace in any literal, which includes this case, and
aligns with the Functional Syntax.

> 4) The BNF for dataAtomic should use a literalList.
> 
> 5) The narrative text in section 2.3 includes nonterminals that are
> not styled as such.
> 
> 7) The BNF for atomic should use an individualList

Fixed

Diffs for 2/, 4/, 5/, and 7/ are
http://www.w3.org/2007/OWL/wiki/index.php?title=ManchesterSyntax&diff=21068&oldid=21067

> 1) Throughout the document the Manchester syntax is described as a
> syntax for "OWL 2 DL ontologies".  It is unclear why it is restricted
> to OWL 2 DL and could not be used to serialize OWL 2 ontologies which
> do not satisfy the DL constraints in SS&FS.

Good idea.  I've done this, and deferred to SS&FS for the conditions in
Section 2.6 which is now dramatically shorter.

Diffs at
http://www.w3.org/2007/OWL/wiki/index.php?title=ManchesterSyntax&diff=21069&oldid=21068 

Note that the Manchester syntax is untyped, and thus needs the OWL 2 DL
typing conditions to disambiguate.  

The diffs for this are
http://www.w3.org/2007/OWL/wiki/index.php?title=ManchesterSyntax&diff=21072&oldid=21071

(I dropped this bit the first time through.)

> 8) The HasKey axiom in the example ontology has description and
> property inverted.
> 
> 10) The syntax would be more consistently styled if HasKey axioms were
> part of a class frame and not in the 'misc' group.

HasKey is now a part of a class frame (which makes it slightly less
general, but see below for more on this).

Diffs, including a couple of fixes to the grammar productions are
http://www.w3.org/2007/OWL/wiki/index.php?title=ManchesterSyntax&diff=21070&oldid=21069

> 11) There are at least two classes of axioms that cannot be expressed
> in the frame syntax as presented:
> a) Class axioms where the first argument is not a named class e.g.,
> SubClassOf( ObjectIntersectionOf(A B) C )
> b) Object property axioms where the first argument is an
> InverseObjectProperty e.g., SubPropertyOf( InverseObjectProperty(p)
> q).  This problem also applies to sub object property chains where the
> super property is an InverseObjectProperty.

Not all OWL 2 ontologies can be directly written in the Manchester
syntax.  This should have been noted in the document.  I have added a
sentence to the introduction and added to the translation in Section
4.3.

Diffs are
http://www.w3.org/2007/OWL/wiki/index.php?title=ManchesterSyntax&diff=21071&oldid=21070 

> -- 
> Mike Smith
> Clark & Parsia

peter
Bell & Labs

Received on Wednesday, 1 April 2009 00:27:32 UTC