If it helps, I thought to call conformance type 1 the 'benchmark'
(it is a sender capabilities minimum requirement for conformance)
and the conformance type 2 the 'ceiling' (it is where the sender
is bound to not send anything which does not conform to a subset
schema (and the subset schema sets the outer bounds of what
can be sent) - in a certain mode (outside of that mode the profile
does not apply but other profiles might such as overall standard
conformance).
I would suggest that these two conformance types - ceiling and
benchmark - could be defined using separate subsets, all within
one 'subset conformance profile'. One part of the conformance
clause could define the benchmark using one subset (e.g. the
subset I labelled just 'subset' and of the order of UBL 1.0 SBS)
and another part of the clause could define the ceiling using
a separate larger subset (obviously it has to be the same 'size'
or larger, such as the subset schema set I called 'subsetPlus').
It would make sense to me if the proposed basic subset for
UBL 2.1 did in fact use a benchmark like my 'subset' sent
previously but defined using XPaths (e.g. in both a table in
the spec and perhaps test assertions based on that table)
and it would make sense if it also included a ceiling like
the subsetPlus schema set I sent where all BIEs in well
known existing subsets such as OIO were included along
with the previous benchmark subset BIEs too. Then a
conforming target implementation would ensure it can
send all BIEs in the 'inner', 'benchmark' and also ensure it
has a mode of use (perhaps indicated using one of the
purpose made UBL IDs like ProfileID or CustomisationID)
where it does not send anything not valid according to the
ceiling. The ceiling conformance could be tested with a
ceiling subset and there might be a need for test cases
(perhaps using the XPaths) to test that every possible
benchmark BIE *CAN* be included in any document.
----
Stephen D Green
On 2 August 2011 14:48, Jon Bosak <> wrote:
Thanks, Stephen. I'm going to arrange a discussion of this in an
upcoming TC call.
Jon
Stephen D Green wrote:
Third time lucky... I'm still getting to grips with this. Sorry
for my slowness to understand it all.
Now I think some more I reckon there can be two main
conformance levels/types (or three if you add one combining
the other two) for a given subset.
conformance type 1: (as I wrote previously)
the sender
can send every BIE in a conformance (sub)set of the subset
the receiver
can receive a document with every BIE in the subset
but then there is the other type, to which I think you allude
conformance type 1: (as I wrote previously)
the sender (has a mode of operation in which it ...)
can send a document with no BIEs which are not in a conformance (sub)set of the subset
the receiver
can receive a document with every BIE in the subset
and
can fail to receive (reject) any document with BIEs which are not in a conformance (sub)set of the subset
The second can make the schema central to conformance tests.
The second is more valuable to receiving systems but may be less testable (not every doc can be tested).
I guess both have pros and cons.
The first is to my thinking more akin with Postel's law putting less onus on senders. Both are valid though.
Best regards
----
Stephen D Green
On 2 August 2011 09:27, Stephen D Green < <mailto:>> wrote:
Hi Jon
Reading through your question again, I may not have answered it.
Are you asking for elaboration on the actual 'predicates' in the
test assertions?
In other words, the actual logic of the suggested profile spec
statements?
I guess they are aimed more at a sending system than at the document
sent.
That is the key to understanding how they differ from (subset)
schema validity.
The schema constrains the document but these statements (and test
assertions
based on the statements) go ufrther by stating that a sending system
MUST
be able to send X, Y and Z in a document. That is not to say a
document MUST
contain X, Y and Z. If Z is mandatory in the schema, one document
might contain Y and Z and another document might contain X and Z. If the system
can send
a document with X and Z and it can send a document with Y and Z
it conforms.
It it cannot send Y and Z then it doesn't conform to this profile
(even though its
documents might all be valid according to the schema). If it sends
documents
with X and Y but not Z then there will be a failed test but this
cannot always be
relied on for conformance because tests cannot prove it will never
send such
a document. This is the weakness of using a schema and validity of
documents
alone as a conformance test - you can never test all documents. My
suggestion
is to apply conformance testing such that the sending system has to
prove that
it can send some set of documents which between them contain all
BIEs in a
set. (It is a little more complex in that the subset itself may be a
superset of all
the BIEs in the set required for conformance. A conforming system
can still
conform, according to a given profile spec, even if it never, say,
implements
each and every, Signature BIE, provided these BIEs are not part of
the set
which determines conformance. The Signature BIEs might all be there
in the
subset schema though.) ----
Stephen D Green
On 1 August 2011 17:49, Jon Bosak <
<mailto:>> wrote:
Hello Stephen,
I like the idea of a basic subset (always have), and I'm happy
to have
the benefit of your work on this.
A question about the test assertion approach, however. I'm having
trouble seeing how these assertions go beyond just saying "you
have to
validate against the subset schema, and in addition, these
optional UBL
elements are mandatory." (In which case, of course, you could just
modify the subset schema to that effect.) Could you help me
understand
this?
Jon
Stephen D Green wrote:
I would refine the previous comment to say that the previous
receiving system
conformance clause might not have been fully testable (since
not every possible
document received will be tested). I could refine this to
say that a receivng system
MUST be able to receive a document which contains every
element in the subset
and every multiple cardinality element twice or more times
in the document.
I would suggest that continuing along the lines of a
conformance test centric subset profile there might be
consideration of test assertions for such a
profile, again considering testability of the assertions and
the normative
statements from which they are derived.
Here is an example of a test assertion set for, say, a
conformance clause relating to
the invoice document type (there might be a clause for each
type to allow
implementations to implement just one or more document types
and still
be conformant).
It is based on an example profile which somehow lists every
element (ignoring
attributes for conformance requirements except insofar as
they are ever mandatory
for the UBL standard schema validity for that document type)
e.g.
Where the profile for the subsets contains statements like
this:
...
Statement INV004: A subset sending system which can send a
subset invoice MUST be able to send a document valid
according to the OASIS standard UBL 2.1 Invoice schema and
containing element /in:Invoice/cbc:ID
Statement INV005: A subset sending system which can send a
subset invoice MUST be able to send a document valid
according to the OASIS standard UBL 2.1 Invoice schema and
containing element /in:Invoice/cbc:CopyIndicator
... one such statement for each element in the subset or a
table to the same effect or some other
blanket statement to this effect including reference to an
overall schema for the subset
...
The conformance clause for subset invoice sending systems
would require conformance to
all these statements. Another clause for receiving systems
might require conformance to
a statement that receiving systes receiving a subset invoice
be able to receive one with all
elements in the subset (including multiple occurences where
the schema includes multiple
occurences).
Test assertions could be presented in the form of markup
like this
-- here using OASIS (in progress) Test Assertion Markup
Language:
<!-- one set of normative statements and corresponding test
assertions for every document type in the subset and for
each of these sets one corresponding conformance clause for
sending systems (for that document type) and one for
recieving systems (for that document type) -->
<testAssertionSet
xmlns="http://docs.oasis-open.__org/ns/tag/taml-201002/
<http://docs.oasis-open.org/ns/tag/taml-201002/>"
setid="ubl-subset.example.ta-__set.1" setname="Invoice Subset">
<common>
...
<!-- need some bindings for prefixes used in the XPath
expressions -->
</common>
...
<testAssertion id="ubl-invoice-subset.__example.ta.s4">
<!-- the actual nomative source might be a table or might
rely on a schema to list all elements in the subset
(attributes too but these might be optional) -->
<normativeSource><__derivedSourceItem documentId="..."
resourceProvenanceId="..." uri="...">A subset sending system
which can send a subset invoice MUST be able to send a
document valid according to the OASIS standard UBL 2.1
Invoice schema and containing element
/in:Invoice/cbc:ID</__derivedSourceItem></__normativeSource>
<target>subset sending system</target>
<prerequisite>can send a subset document valid according to
the OASIS standard UBL 2.1 Invoice schema</prerequisite>
<predicate>can send a document valid according to the OASIS
standard UBL 2.1 Invoice schema and containing element
/in:Invoice/cbc:ID</predicate>
<prescription level="mandatory"/>
</testAssertion>
<testAssertion id="ubl-invoice-subset.__example.ta.s5">
<normativeSource><__derivedSourceItem documentId="..."
resourceProvenanceId="..."
uri="...">...</__derivedSourceItem></__normativeSource>
<target>subset sending system</target>
<prerequisite>can send a subset document valid according to
the OASIS standard UBL 2.1 Invoice schema</prerequisite>
<predicate>can send a document valid according to the OASIS
standard UBL 2.1 Invoice schema and containing element
/in:Invoice/cbc:CopyIndicator<__/predicate>
<prescription level="mandatory"/>
</testAssertion>
<testAssertion id="ubl-invoice-subset.__example.ta.s6">
<normativeSource><__derivedSourceItem documentId="..."
resourceProvenanceId="..."
uri="...">...</__derivedSourceItem></__normativeSource>
<target>subset sending system</target>
<prerequisite>can send a subset document valid according to
the OASIS standard UBL 2.1 Invoice schema</prerequisite>
<predicate>can send a document valid according to the OASIS
standard UBL 2.1 Invoice schema and containing element
/in:Invoice/cbc:UUID</__predicate>
<prescription level="mandatory"/>
</testAssertion>
<!-- one normative statement and corresponding test
assertion for every element in the subset for that document
type -->
...
<testAssertion id="ubl-invoice-subset.__example.ta.r1">
<normativeSource><__derivedSourceItem documentId="..."
resourceProvenanceId="..."
uri="...">...</__derivedSourceItem></__normativeSource>
<target>subset receiving system</target>
<prerequisite>can receive a subset document valid according
to the OASIS standard UBL 2.1 Invoice schema</prerequisite>
<predicate>can receive a document valid according to the
OASIS standard UBL 2.1 Invoice schema and containing every
single-occurence element in that subset once and
multiple-occurence element in that document twice or
more</predicate>
<prescription level="mandatory"/>
</testAssertion>
...
</testAssertionSet>
Best regards
----
Stephen D Green
On 30 July 2011 18:40, Stephen D Green
<
<mailto:>
<mailto:stephengreenubl@gmail.__com
<mailto:>>> wrote:
Suggestion for a basis for subset profile conformance clauses
In the UBL 1.0 Small Business Subset were
clauses which
I found later to be very difficult to test precisely and this
might make conformance testing and conformance claims
problematic. The introduction of conformance clauses in
recent standard specifications is to aid interoperability and
promote adoption through the clarification of what it means
for an implementation to conform to a standard specification.
To promote adoption of subsets for the OASIS Universal
Business Language it is important to include a conformance
clause of set of clauses and I would suggest that a subset
conformance clause should target more than just the UBL
documents (Invoice, Order, etc) themselves but also there
should be a conformance clause for a sending system, one
for a receiving system instead of or in addition to the
clause
for conformance of the documents themselves. I would like
to suggest as a basis for the clause for the conformance
target of sending system conformance to a set of statements
which amount to the sending system being able to send a
certain set of BIEs in a given document, as defined by the
subset schema for that document. I would suggest that a
basis for a clause for the target of a receiving system
would be a conformance clause requiring that the system MUST
NOT reject a document merely because of the presence in
it of any of the BIEs as defined by the subset schema (or
list of BIEs, e.g. given as a set of XPath expressions).
As a simplistic example, if a subset contains BIEs
X,Y,Z of
which Y and Z are mandatory and X is optional, there could
be a set of specification requirements to the effect that the
sender system MUST be able to send all BIEs X, Y and Z
in sending that particular document (even though only Y
and Z are mandatory in any given document of that type).
A conformance clause for the target of the sending system
would mandate these particular statements as necessary
for the conformance of that system. A set of statements that
target the receiving system would require that the system
MUST NOT reject a document of that type merely because
it contains BIE X or Y or Z. A conformance clause for the
receiving system would make the statements mandatory
for conformance by such a system. There would be a set of
statements for each conformance target that the documents
of that type MUST contain BIEs Y and Z because these are
mandatory.
Such conformance clauses and this focus on
testability may
help to promote adoption of any given subset and contribute
to successful adoption and interoperable implementations.
Best regards
----
Stephen D Green
