Re: Testable assertion tagging for W3C specifications

On Tue, 28 May 2002, Dimitris Dimitriadis wrote:

> Also, but this is a matter of discussion, I take the issue of
> being able to adress a spec to be one special case of being able
> to adress at all, something which we can do from tests too. Why
> not achieve uniformity, in case it simplifies and gives rise to
> added value?

If uniform addressing and query scheme "simplifies and adds value", it
should be adopted!

I am convinced there is no such thing as a simple uniform addressing
and query scheme. The problem domain is just too broad and complex. If
you think otherwise, consider supporting SQL-like queries in your
scheme. Once you are done, consider supporting queries that SQL is
known to support poorly. If you argue that supporting SQL-like queries
is not needed, you begin to lose uniformity; some examples on this
thread would be easy to implement using SQL and a relational database
of test assertions/rules.

Yes, addressing a test rule is a special case of addressing. And
addressing is a special case of a "select" query. And a select query
is a special case of a query. Etc., etc. Simple solutions become
possible only when you limit your problem domain enough. I believe we
can come up with a simple addressing scheme if we limit the problem
domain to, say, the extraction of single pieces from XML documents.
Perhaps such a simple addressing scheme already exists. Moreover, I
would not be surprised if such domain limitation will cover most of
existing basic needs; with more complex, custom solutions covering the
rest.

I would suggest that we collect the requirements for "Testable
assertion tagging for W3C specifications" (the subject of this
thread). That is, what exactly do people need to be tagged or
addressed? Once we have the requirements, we can decide on the problem
domain and on the addressing/query scheme.

I can contribute my own requirement, derived from working on an HTTP
test suite:

	- an address must be able to uniquely identify any continuous 
	  text within HTTP RFC 2616 document

I think the above will cover all basic needs of my test suite,
including:

	- pointing to a single MUST/SHOULD/MAY requirement 
	  (using a single address)

	- pointing to a set of assertions related to a single 
	  test case (using a set of addresses stored with/in
	  a test case object)

	- searching for test cases covering a given MUST/SHOULD/MAY
	  (using a reversed index of addresses)

	- searching for test cases with a given known-in-advance
	  property (using an index of test cases; not related
	  to the spec-addressing scheme since these queries
	  are domain-specific!)


Thanks,

Alex.

Received on Tuesday, 28 May 2002 18:21:10 UTC