W3C home > Mailing lists > Public > www-dom-ts@w3.org > April 2001

Re: [General] Language-independent test representation

From: Joseph Kesselman <keshlam@us.ibm.com>
Date: Fri, 20 Apr 2001 11:42:50 -0400
To: "'www-dom-ts@w3.org'" <www-dom-ts@w3.org>
Message-ID: <OF85E4B80A.228E87DB-ON85256A34.005593E9@pok.ibm.com>

My one hesitation with this proposal is that, when I've tried to craft
nontrivial testcase scripts in XML in the past, I've found myself either
(a) unrolling any logic in the test to turn it into a flat series of
operations, which makes the script both huge and hard to read, or (b)
trying to write a fairly complete programming language in XML syntax, in
order to handle the conditionals/logic/variables... which snowballs rapidly
into a much larger project than the tests themselves.

If we assume that programming languages are inherently fairly regular, it
may be possible to write a simple XML-based pseudocode language which can
be "blindly" translated into specific languages. But I suspect that larger
tests are likely to push the boudaries of this approach as they start to
require non-DOM programming activity to generate/manuipulate/validate the
information... and it does mean an ongoing effort to maintain and develop
the test language and its mappings.

It's definitely worth looking at. But I won't be surprised or terribly
disappointed if it turns out that human porting of many (most?) testcases
is needed after all. If programming languages were really that similar, we
wouldn't generate quite so many of them.

Joe Kesselman  / IBM Research
Received on Friday, 20 April 2001 11:43:35 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:34:02 UTC