RE: interoperability is... ? (was: SML interop test plan)

Funny how "clear" terms become fuzzy once you try to grab them by the 
tail.  My brain thinks interoperability is a continuum, and the degree to 
which it is achieved == the degree to which two independently developed 
implementations extract the same semantic meaning from the same input, for 
each feature supported.  In this way of thinking, implementations (not 
artifacts like documents or models) are more or less interoperable.
I freely admit that I have used "widely interoperable" about models 
containing sml:uri, and my usage is in conflict w/ the "def" above.  A 
more deliberate wording for that might be: referentially conforming SML-IF 
documents are likely to have a larger number of implementations able to 
process them.
My def above is consistent I think with what Kirk cited in the interop 
test plan.  Either way you slice it, a sophisticated view of interop 
entails understanding the set of features supported by the implementations 
involved, intersected with those actually used by the models involved.

Best Regards, John

Street address: 2455 South Road, P328 Poughkeepsie, NY USA 12601
Voice: 1+845-435-9470      Fax: 1+845-432-9787



From:
"Wilson, Kirk D" <Kirk.Wilson@ca.com>
To:
"Kumar Pandit" <kumarp@windows.microsoft.com>, "Smith, Virginia (HP 
Software)" <virginia.smith@hp.com>, <public-sml@w3.org>
Date:
04/17/2008 01:56 PM
Subject:
RE: SML interop test plan



This is a good start on a difficult topic to think about.  Here’s a wild 
question I have on lines 37-38 (in section 5):
 
For the purpose of this test plan, two implementations are said to be 
interoperable if they produce identical model validation result for each 
test case that tests a required feature of SML and SML-IF.
 
What is the primary object of interoperability?  Are “implementations” 
said to be interoperable (only once implied in the SML-IF spec), or are 
interchange models “interoperable” (which seems the dominant view in the 
SML-IF spec, see section 4.5), or is an SML-IF document “interoperable” 
(also in the SML-IF spec, and SML-IF “document” is close enough to 
“interchange model” to make this case essential equivalent to the 
immediately preceding case).  Are we defining a new sense of 
“implementation interoperability” for purposes of the test plan?
 
This sentence above seems to imply that two implementations of SML model 
validators can be said to be interoperable they have produce identical 
model validation results…etc.  I don’t believe the specification ever 
defines interoperability of SML model validators except via the 
interoperability of an interchanged model.  (Indeed, the SML-IF spec 
explicitly that SML model validation permits “points of variability” on 
which SML-IF imposes additional constraints to address.)  In other words, 
interoperability for models is defined between an SML-IF producer and 
SML-IF consumer.  A model is interoperable if an SML-IF model validator 
assesses the interoperability of the interchange model in a way that 
corresponds to the way the producer “thinks” the interoperability should 
work.
 
To take an extreme example:  My SML-IF producer understands an EPR scheme 
and so builds an SML-IF document with the appropriate aliases and embedded 
documents to handle my EPR SML references.  But your SML-IF validator 
doesn’t understand these aliases (e.g., it doesn’t understand the EPR 
Reference Scheme I’m using), so the model is not interoperable between the 
two implementations.  In a derivative sense, the implementations are not 
“interoperable”.  If this is an accurate picture of the situation, then 
the fact that implementation interoperability is “derivative” might be 
clarified in the test plan.
 
Kirk Wilson, Ph.D.
Research Staff Member, CA Labs
603 823-7146 (preferred)
Cell:  603 991-8873
 
This e-mail may contain information that is privileged or confidential. If 
you are not the intended recipient of this communication, please delete 
this e-mail and notify the sender immediately.

From: public-sml-request@w3.org [mailto:public-sml-request@w3.org] On 
Behalf Of Kumar Pandit
Sent: Tuesday, April 15, 2008 10:18 PM
To: Smith, Virginia (HP Software); public-sml@w3.org
Cc: Kumar Pandit
Subject: RE: SML interop test plan
 
Good suggestions for numbering. Here is the updated pdf with 
section/page/line numbers. I have also attached the source word document.
 
I agree about not needing to duplicate the IF format since we can readily 
use IF itself. The support for locator element is optional. Not all 
implementations may support it which makes it hard to use for interop 
testing. However, the bigger problem with multiple files is that of change 
management. Change to one file may affect many tests and thus each person 
making a change will need to run all implementations against all tests to 
ensure no breaks. This is one of the reasons why the document recommends a 
single SML-IF file per test case.
 
 
From: public-sml-request@w3.org [mailto:public-sml-request@w3.org] On 
Behalf Of Smith, Virginia (HP Software)
Sent: Tuesday, April 15, 2008 3:48 PM
To: public-sml@w3.org
Subject: RE: SML interop test plan
 
Nice work on the doc, Kumar. 
 
I have a question on the test-description file – 1b. It seems like you 
wouldn’t have to recreate the IF format – you could just use the IF format 
for the test harness, probably with the locator element. So the test 
harness would just have to embed the appropriate documents into a final IF 
file to run the test. And 2d would still apply here.
 
I also recommend adding page number and line numbers for easy reference - 
especially if you are not providing a word doc that can be marked up.
 
--
ginny
 
 
From: public-sml-request@w3.org [mailto:public-sml-request@w3.org] On 
Behalf Of Kumar Pandit
Sent: Thursday, April 10, 2008 9:31 PM
To: public-sml@w3.org
Cc: Kumar Pandit
Subject: SML interop test plan
 
Here is the first draft of SML interop test plan. Please review and send 
your feedback.
 
I have not included the list of tests because I need to make some updates. 
I will send it later.