Minutes from February 4 conference call

["Michael Kass" <michael.kass@nist.gov>]

Title: call this Monday

To all,

 

   Attached are the minutes from the last IIC conference call.  Jacques has changed his email address (to jdurand@us.fujitsu.com )

and could not yet post to the IIC mailing list.

 

Cheers,

Mike

 

IIC Conference Call Minutes: 
Monday, February 2, 2004
 
Attendance:

Mike Kass (NIST)
Jacques Durand (Fujitsu)
Serm Kulvatunyou (NIST)
Monica Martin (Sun)
Steve Yung (Sun)

Excused:
Tim Sakach (DrakeC)
Pete Wenzel (SeeBeyond)


Minutes taker: Jacques Durand

Time: Monday February 2, 2004, 2pm PT
Host: Fujitsu
Toll - :  1-512-225-3050

Agenda: 

1. Test Framework upgrade: (Mike Kass)
- finalize test cases control flow issues (concurrent threads, fork/branching, exceptions) 
- review of TestFramework draft, remaining issues.
- position on standard interface/API Test Service - MSH (see notes Jan 5th)

2. Applying the test framework:
- IIC Test plan proposal for ETSI European test event: discussion on
"testbed contributions" from vendors. 

3. Others
- "light" IIC f-2-f New Orleans April 28-29 details.

--------------------------

1. TestFramework status:

Two main issues being discussed:
- control flow for test cases.
- API TestService-specUnderTest, and TestService-TestDriver.

Control flow:
- do we need a "sleep" operation? (either as independt step, or a delay op for any step)
Yes. We need to manipulate timing of any step. BPSS has time limits on entire transactions,
that will require delay simulation for testing.
- converging on concurrency management ops (split / join) which will comply with
usual wflow  (and BPSS) semantics. Split will always start async threads.
- It appears we don't need "OR split" (fork+XOR in BPSS) as the test driver 
always decides of which thread(s) to start (not an app behind). 
If one of alternative threads may be started by other party, we can start them all 
and or-join.
- Monica: can relay any question we have for ebBP TC. The TC can also review
our design.

API:
- goal would be to make it easier to candidate implementations under test to
interface with any test framework (wrapper code would be same or very similar.)
- Mike already proposed a simple one-call abstract API (3.2.3 of TFramework). 
- possible issue with the "envelope" given as argument, when T.Service sends a 
message: this envelope shold be an abstract, format-neutral declaration of 
header attributes.

Other issues:
- Persistent message store management is unclear: "GetMessage" is NOT removing
read messages, so two consecutive GetMessage ops will see the same messages
(idempotent). This is contrary to the expected behavior in general: by default GetMessage
should "consume" messages so that they are not visible anymore to subsequent ops.
- when GeMessage selects several messages, what is the semantics of the assertion
in the rest of test step: which message is concerned?


2. Applying the test framework:

- Test plan posted on our site. Still a draft, misses detail of test cases.
Korbit has implemented about 100 conformance test cases.
- Korbit will be main testbed provider. Considering involving other providers:
Mike said that NIST testbed would NOT be ready in time as an alternative.
DrakeCertivo  willing to consider some contribution.
- ETSI will do announce soon.

3. Others
- We'll have a "light" IIC f-2-f New Orleans April 28-29, just a small meeting
at convenient time, as there will be conflicts with other meetings.
Schedule unknown yet, but will be sometime over Wed-Thu.