tag-comment message
[Date Prev]
| [Thread Prev]
| [Thread Next]
| [Date Next]
--
[Date Index]
| [Thread Index]
| [List Home]
Subject: RE: [tag-comment] Re: [tag] coments on Section 3.7.2 "COmposition of Assertions" (V0.995)
- From: "Durand, Jacques R." <JDurand@us.fujitsu.com>
- To: <david_marston@us.ibm.com>, <tag-comment@lists.oasis-open.org>
- Date: Thu, 23 Oct 2008 09:47:27 -0700
David:
>I
see a problem when a prerequisite requires testing in order to determine
eligibility.
>One can then build a dependency chain of the
tests: e.g., you can't apply tests derived from TA 76345 until
>you have an outcome on any of a family of
tests (list them here) that will show that the prerequisite of 76345 has been
satisfied.
>Some of those tests may have their own
prerequisites and hence dependencies, etc.
Yes but that is
precisely the behavior expected by some TA users:
- in many cases eligibility requires anyway some
"testing" i.e. when the prerequisite is itself a predicate
expression.
- in some of these cases, eligibility refers to another
TA(s) - meaning to the outcome of the predicate of these TAs. It is more
informative to "reuse" another TA in the prerequiiste Instead of just
replicating the predicate expression: this indicates to test suite writers a way
to avoid redundant testing.
It also indicates what is the impact of failing a TA on
the outcome of other TAs (so that users can make the difference between a
"naturally unqualified" target and a target that is not qualified because
it failed a previous TA where it should have passed
it.
Yes that introduces some partial order among TAs
(dependencies) but again thats what some users need (e.g.
WS-I).
You do not have to use it
:-)
Jacques
I see a problem when a prerequisite
requires testing in order to determine eligibility. One can then build a
dependency chain of the tests: e.g., you can't apply tests derived from TA 76345
until you have an outcome on any of a family of tests (list them here) that will
show that the prerequisite of 76345 has been satisfied. Some of those tests may
have their own prerequisites and hence dependencies, etc.
Prerequisite: [the widget] is conformant to
Mini-Widget Small Box Specification 1.2
Predicate: [the widget] is conformant to WidgetSpec 1.0
<SG> the main
assertion is that such a small widget is also a widget in that it conforms to
the WidgetSpec 1.0
the way this TA reads, the normative Source must be aligned with the
Predicate: it should be "Conformance clause to
WidgetSpec 1.0" (exactly like in widget-TA109-1)
I think it reads that you must already
determine that the target is a Small Box 1.2 Mini-Widget before you can even ask
about whether it is a 1.0 Widget. Circular dependency?
<SG> Surely this is just another way of saying the same
thing:
"conformant to WidgetSpec 1.0" == "conforms to
the conformance clause of WidgetSpec 1.0"
That equivalence is true if WidgetSpec 1.0 only has one class
of product. Otherwise, it must say something like "is a conformant [class] as
defined by WidgetSpec 1.0".
The presence of the prerequisite means that the TA only
applies to widgets already known as confirming to Mini-Widget Small Box spec:
<SG> I agree
The problem is the "already known" part. I think this drives
the debate about whether prerequisites can be based on either (1) declarations
from the product provider and (2) measurable/testable properties of the
implementation. A "type 2" prerequisite, if as broad as "is a conformant widget"
could still exist in a formal structure of measured facts (predicates, as it
were) as long as the TAs for widgets include a final "summary" TA for overall
conformance.
I think you
can see why I wish this example used an orthogonal spec. The example I suggested
is power sources for portable devices. Specify that some but not all widgets are
portable, and certain portable ones must conform to the specs about power
sources.
.................David
Marston
IBM Research
[Date Prev]
| [Thread Prev]
| [Thread Next]
| [Date Next]
--
[Date Index]
| [Thread Index]
| [List Home]